Fragments, Not Intact Genomes, Are What Survive

Ancient DNA works because laboratories treat every fragment as suspect, combining dense substrates, clean room control, damage patterns and statistical comparison, so a result is trusted only when contamination would have to imitate several independent signs of age at once. The material recovered from an old bone or tooth bears little resemblance to the long, ordered molecule taken from a living cheek swab. Ancient fragments commonly measure 30 to 70 base pairs long, a length closer to a sentence torn from a book than to the book itself. Millions of such pieces may survive in a single gram of bone, yet each piece ends abruptly, carries chemical changes along its length, and arrives mixed with genetic material from soil bacteria, fungi, plants, animals, and every person who handled the specimen after excavation. The work therefore begins with suspicion rather than celebration, and that suspicion shapes every later choice, from the bone selected for drilling to the statistical test applied after sequencing.

The scale of that mixture explains why early success attracted early doubt. In 1984, Higuchi and colleagues reported sequences recovered from a dried museum skin of the quagga, an extinct relative of the zebra, in work often credited as the first published recovery of genetic material from an organism no longer alive. A year later, a 1985 report by Svante Paabo described sequences attributed to an Egyptian mummy, a claim later treated cautiously because early copying methods could amplify a single modern contaminant as efficiently as an authentic old molecule, and because the checks needed to separate the two had not yet matured. The contrast between those two episodes taught the field a durable lesson. Preservation was possible, but possibility alone proved nothing about any particular result, and a sequence matching a plausible ancestor could still have entered the tube long after burial.

For that reason, success in the discipline rarely means reading a complete genome from a single bone in a single pass. It usually means assembling overlapping fragments recovered across many extraction attempts, comparing them against reference genomes, and asking whether the pattern of agreement and disagreement holds together under independent testing. The inference rests on accumulation and on consistency, and both depend on excluding pieces that do not belong to the original vessel at all.

That exclusion problem never disappears, because contamination mimics the target in the most dangerous way possible. The discipline answers that risk by refusing to let any one sign carry a conclusion. Work takes place in rooms designed to keep modern material out. Damage patterns expected in old molecules are measured rather than assumed. Replicate extractions are compared, negative controls are sequenced alongside real samples, and results are tested against population patterns that contamination would struggle to reproduce. A finding earns confidence when several unrelated checks point the same way, and it loses confidence the moment one check requires special pleading to explain.

Understood in that spirit, the subject is less a treasure hunt for intact genomes than a forensic exercise conducted at molecular scale. Method turns those traces into evidence by showing, step by step, that the traces behave like old, damaged, endogenous material and unlike anything introduced in the field, the museum, the store room, or the laboratory itself.

The Question Paleogenomics Is Trying to Answer

Paleogenomics asks who was related to whom, which populations moved and mixed, and how biological inheritance changed across time, using genetic material recovered directly from past organisms rather than inferred only from living descendants. The distinction matters because living populations are survivors, and survivors give a filtered account of the past. A lineage that left no living descendants disappears from a survey built only on present day samples, however large that lineage once was. Direct sampling through time repairs those blind spots by placing individuals at known points in sequence, so change itself becomes observable rather than reconstructed backward from a single endpoint.

The power of that arrangement appeared in stages. In 1997, Krings and colleagues recovered the first Neanderthal mitochondrial sequence, taken from the Neanderthal type specimen found in the Neander Valley, and compared it with corresponding sequences from living humans. The Neanderthal sequence fell outside the range of variation seen in the living sample, a result that supported a deep separation between the two lineages while still leaving open the separate question of whether the two had interbred. Mitochondrial inheritance follows the maternal line only and represents a small fraction of total inheritance, so a single locus could never settle that larger question. A broader answer arrived when nuclear genomes became readable. Green and colleagues reported in 2010 that the Neanderthal genome comparison supported an estimate of Neanderthal ancestry between 1 and 4 percent for genomes outside Africa, indicating that interbreeding between Neanderthals and the ancestors of many living populations had occurred after those populations left Africa. The mitochondrial result and the nuclear result did not contradict each other. They measured different inheritances, and the difference between them illustrated a central habit of the field. Every claim is bounded by the part of the genome actually recovered, the individuals actually sampled, and the comparison actually made, and a careful claim narrows as evidence widens rather than swelling to fill an entire question at once.

The same direct sampling extends well beyond human relatives. In each case the gain is temporal depth. A single time slice can show pattern, but a series of time slices can show direction, speed, and reversal.

None of that reach removes the limits imposed by preservation and sampling. Remains cluster where conditions favoured survival and where excavation has been intensive, so the map of recovered sequences reflects geology and research history as much as past population density. Relatedness estimates also depend on reference genomes and statistical models, each carrying assumptions about population size, mutation, and mixture that must be stated rather than hidden. The discipline therefore treats every reconstruction as provisional in a specific sense. New samples from an unsampled region or period can revise a migration model substantially, not because earlier work was careless, but because earlier work described only the individuals that preservation happened to deliver.

The question, then, is comparative and historical rather than merely identificatory. Naming the species of a bone is a beginning, not an achievement. The achievement lies in placing that bone inside a network of relatives, movements, mixtures, and extinctions, and in showing how that placement changes when material from another century, another valley, or another branch of the same family enters the comparison.

How DNA Breaks Down After Death

Death removes the maintenance system that kept the molecule intact during life. A living cell detects and repairs chemical damage continuously, replacing broken bases, mending single strand breaks, and discarding structures too damaged to use. Once circulation and metabolism stop, repair stops with them, while water, oxygen, enzymes released from the dying cells themselves, and microbes feeding on the tissue keep working on the exposed material. Decay is therefore not a single event that happens to a genome at burial. It is a long accumulation of chemical insults, each one shortening, altering, or crosslinking fragments until the survivors represent only the protected minority that escaped the worst conditions for the longest time.

Hydrolysis drives much of the early fragmentation. Water attacks the bonds linking bases to the sugar phosphate backbone, and loss of a base leaves a weakened site where the strand breaks more readily under later stress. Enzymes from the organism and from colonising bacteria cut the molecule directly, which is one reason soft tissue rich in such enzymes loses recoverable material far faster than dense mineralised tissue does. The result, repeated over centuries, matches the observed reality of ancient fragments commonly 30 to 70 base pairs long, with longer pieces becoming rarer as age and exposure increase. Fragment length itself thus becomes evidence. A sample dominated by very long, undamaged molecules in a context that should have broken them down invites suspicion rather than confidence.

Chemical alteration proceeds alongside physical breakage, and one alteration matters more than any other for authentication. Cytosine, one of the four bases, converts over time into uracil, a base normally associated with a different nucleic acid, in a process called deamination that happens fastest near the exposed ends of a fragment. When a copying enzyme later reads a uracil as if it were thymine, the sequenced copy shows a characteristic substitution, recorded as a cytosine to thymine change at fragment ends, with a complementary pattern at the opposite end of the same piece. The pattern is uneven, concentrated toward the termini, and correlated with age and preservation conditions, which makes it hard for a modern contaminant to reproduce by accident. Laboratories measure that terminal damage directly, use enzymes that remove uracils when a cleaner interior sequence is wanted, or retain damaged molecules deliberately when the damage itself forms part of the argument for authenticity. Either way, alteration that looks like corruption from a purely informational standpoint becomes one of the strongest tools the field possesses.

Rate governs all of it, and rate depends overwhelmingly on temperature. Chemical reactions that break bonds and deaminate bases proceed faster in warmth and slower in cold, so a decade in a hot, wet soil can damage a molecule more thoroughly than a millennium in frozen ground. Moisture matters because hydrolysis requires water and because wet ground supports the microbial communities that consume soft tissue and attack exposed bone. Soil acidity dissolves the mineral that shelters molecules inside bone and tooth, while stable, dry, neutral conditions slow nearly every destructive pathway at once. No single clock therefore measures survival across all sites. Age alone predicts little without the thermal and chemical history that accompanied it, which is why two bones of identical age, buried a short distance apart in different soils, can differ entirely in what either one still contains.

Where Ancient DNA Survives Best

Survival favours places where cold, dryness, stability, and mineral protection coincide, and it punishes places where heat, water movement, and acidity work together for long periods. Permafrost offers the most forgiving combination known, because sustained freezing slows chemical decay, limits liquid water, and suppresses microbial activity for as long as the ground stays frozen. Dry caves rank close behind when their interior temperature stays low and steady, their sediment stays undisturbed, and percolating water does not repeatedly wet the buried material. Arid shelters, high altitude sites, and sealed contexts inside dense sediment can also preserve well, while open tropical soils, flooded ground, and acidic forest floors destroy material so thoroughly that even comparatively recent remains may yield nothing usable.

Substrate matters as much as setting. Bone is porous and varies widely, tooth enamel protects the dentine and pulp chamber beneath it, and one particular bone outperforms the rest by a wide margin. The petrous portion of the temporal bone, the dense pyramid of bone surrounding the inner ear, is the hardest substantial structure in the human skeleton, and its density limits water movement, microbial invasion, and mechanical damage around the material trapped inside it. Gamba and colleagues demonstrated in 2014 that the petrous portion yielded substantially more endogenous material than other skeletal elements from the same individuals, and Pinhasi and colleagues extended and systematised that comparison in 2015, confirming the inner ear region as the preferred sampling target where preservation of the skull permits its use.

How long can DNA survive in bone?

DNA can survive in bone for hundreds of thousands of years under cold, dry, stable conditions, and exceptional permafrost material has yielded sequences approaching one million years or more, although survival depends less on age alone than on temperature, moisture, chemistry, and protection from repeated thawing and microbial attack.

Van der Valk and colleagues reported in Nature in 2021 that mammoth remains from Siberian permafrost yielded genomes between roughly one million and 1.65 million years old, recovered from teeth whose mineral structure had sheltered fragments through an immense span of frozen stability. Most bone from most places fails far earlier, and failure rates rise steeply outside cold contexts. A temperate site a few thousand years old may produce excellent material from petrous bone and teeth while producing nothing from ribs or long bones of the same individual, and a tropical site of the same age may produce nothing from any element at all.

Preservation also varies inside a single skeleton and inside a single bone. The outer surface weathers, cracks, and exchanges material with soil, while interior regions stay comparatively sealed. Rootlets, insect activity, and burrowing disturb some deposits and leave neighbouring deposits untouched. Excavation records that note depth, sediment type, water exposure, and association therefore carry real analytical weight later, because a laboratory interpreting a weak yield needs to know whether weak preservation was predictable from context or whether the result contradicts the burial history and demands another explanation. Selection stays selective for ethical reasons as well. Sampling destroys a small part of an irreplaceable specimen, so the best substrate, taken in the smallest sufficient quantity, from the least visually and scientifically damaging location, represents both good chemistry and good stewardship of a finite archive that future methods may read more completely.

Sediment DNA and Genetic Evidence Without a Skeleton

A skeleton is not the only container that burial leaves behind. Genetic material enters cave floors, lake beds, and occupation deposits from skin, hair, faeces, urine, blood, and decaying soft tissue, then binds to mineral particles in the sediment and can persist there long after every visible trace of its source has disappeared. That persistence turns sediment into an archive in its own right, one that records presence rather than anatomy. A sequence recovered from a layer cannot show the shape of a jaw or the wear on a tooth, but it can show that a particular lineage occupied a place during the period that layer accumulated, including periods for which no bone or tooth from that lineage has been found, identified, or preserved in recognisable form.

The demonstration that changed practice came from Denisova Cave in Siberia. Slon and colleagues reported in Science in 2017 that mitochondrial sequences from Neanderthals and Denisovans had been recovered directly from cave sediment samples, including layers that contained no identified skeletal remains of either group. Denisova Cave already held unusual importance because Denisovans themselves had been recognised first from a finger bone sequence rather than from a distinctive skeleton, so the site had a history of genetic evidence outrunning anatomical evidence. Sediment recovery extended that logic from rare bones to ordinary dirt, sampled layer by layer, and matched against the established stratigraphy of the cave. Where a sequence appeared in one layer and vanished from the layers above it, the sediment supplied a rough chronology of occupation that bone finds alone, sparse and unevenly distributed, could not have provided at comparable resolution.

Sediment evidence reaches its greatest age in frozen ground. Kjaer and colleagues reported in Nature in 2022 that environmental material from Greenland sediment, about two million years old, preserved a broad record of the ecosystem that produced it, including plants, animals, and microbial life from a far northern landscape very different from the ice covered ground covering the site in later periods. The result widened the plausible reach of the entire method, showing that under exceptional stability the archive extends to depths of time once considered closed to genetic inquiry of any kind.

Method shapes what sediment can honestly say. Material moves in soil more readily than it moves inside a sealed bone, through water percolation, animal burrowing, root growth, and the digging of later occupants into earlier layers, so a sequence found at a depth may have travelled there from a younger deposit above it. Rigorous work therefore samples sediment in controlled columns, tests multiple layers, compares results against dated material from the same layers, and treats a single positive sample as a prompt for replication rather than a conclusion. The strongest sediment arguments are accordingly modest in wording and strong in structure, resting on repeated recovery across layers, agreement with independent dating, and absence of the same sequence from control samples taken outside the deposit.

The value of the approach lies in coverage. Excavation recovers bones unevenly, museums hold skeletons unevenly, and destructive sampling of rare remains faces justified restraint, while sediment can be sampled widely, repeatedly, and in small quantities from many layers of many sites.

Excavation and Sampling Without Ruining the Evidence

Contamination begins at the moment of exposure, so careful recovery starts before a specimen reaches any laboratory. A bone uncovered by a trowel passes through air carrying skin cells and hair from the excavation team, rests on surfaces touched by many hands, and may be washed, brushed, labelled, bagged, and stored beside material from other contexts before anyone considers its molecular value. Field practice adapted once that cost became clear. Specimens intended for molecular work are lifted with gloves, handled as little as possible, left unwashed when washing would drive surface material deeper into porous bone, packaged separately in clean containers, and documented in place so that context survives even when the specimen itself travels far from the site.

Documentation carries as much weight as physical care. Precise recording of layer, depth, associated objects, sediment type, and position allows later workers to test whether a genetic result fits the burial it came from or contradicts it, and contradiction is valuable information rather than embarrassment. Photography before lifting, separate numbering for subsamples, and a written chain of custody from trench to store room to laboratory bench all belong to the same discipline. The archive a laboratory receives consists of records as well as bone, and gaps in the records narrow what any result, however clean the sequencing, can responsibly claim about the past.

Sampling itself is a controlled act of destruction and is planned as one. The preferred target, often petrous bone or a tooth, is selected after weighing expected yield against the loss involved, and the quantity removed is kept to the minimum that established yields justify. Drilling takes place slowly to limit heat, because friction heat damages the very molecules the drilling aims to recover, and powder is collected from interior regions after the outer surface is removed or decontaminated, since the surface carries the heaviest load of handling contamination and soil derived material. Where a tooth is sampled, cementum and dentine from the root often provide the target while the crown, the most visible and diagnostic part, stays intact. Every cut is photographed, measured, and recorded against the specimen inventory, so the intervention itself becomes part of the permanent record available to anyone re-examining the specimen or the result in later years.

Storage and transport extend the same logic across time. Cool, dry, stable storage slows the decay processes that burial conditions had already slowed, while repeated warming, damp storage, and handling without gloves steadily add damage and contamination to material that survived centuries underground. Restraint governs the whole sequence because the archive is finite and questions change. Material removed carelessly cannot be restored, and material consumed by an early, low yield method cannot be resampled by a later, better one. Permissions from curators, descendant communities, and relevant authorities therefore precede sampling rather than follow it, and a proportion of suitable material is deliberately left unsampled wherever a specimen is unique. That conservation ethic is not separate from the science. A field that reads evidence measured in fragments of 30 to 70 base pairs, extracted from a few hundred milligrams of powder, succeeds precisely by treating every gram of surviving material as evidence that must still exist, in documented condition, after the present question has been answered. The association logic behind that care is the same logic used in other dating problems in archaeology, including how archaeologists date ancient shipwrecks from cargo, coins and residues after a hull has vanished.

Inside a Clean Laboratory Built to Exclude Modern DNA

A dedicated laboratory for old material is organised around a single asymmetry. Modern genetic material is abundant, intact, and everywhere, carried on skin, hair, breath, clothing, and dust, while the target material is scarce, damaged, and present only inside the specimen. Ordinary laboratory practice, designed for plentiful modern samples, would overwhelm the target within minutes. The dedicated facility therefore separates the work physically from any space where modern material or amplified copies are handled, enforces movement in one direction only, from the cleanest stages toward later stages, and never back, and treats every surface, reagent, and item of clothing as a potential carrier until cleaning or testing shows otherwise.

The room itself does much of the work. Air handling maintains positive pressure and filtered supply so that particles move outward rather than inward when a door opens. Surfaces are smooth, minimal, and cleaned with bleach based solutions that destroy the molecule rather than merely moving it around, followed in many facilities by ultraviolet irradiation during unoccupied hours, which damages exposed material on benches and equipment. Workers wear full protective clothing, including gloves, sleeves, hair covering, and face protection, change that clothing on entry, and follow rules about cosmetics, illness, and prior contact with amplified material that would seem excessive in almost any other laboratory setting. Reagents arrive sealed, are tested before use, and are opened only inside the facility. Equipment stays dedicated to the facility rather than circulating through general laboratory space, because a shared centrifuge or pipette can carry enough modern material to ruin a run of low yield extractions.

Controls convert that discipline from intention into measurement. Every batch of extractions includes blank controls that contain no specimen at all and pass through every step the real samples pass through, so any sequence appearing in a blank identifies contamination entering during processing and can invalidate or qualify the batch. The logic is cumulative and unsparing. A result that survives blanks, replicates, and staff comparison has passed tests a contaminant would usually fail at least once, while a result that appears in one batch only, or shadows a blank, is set aside however interesting its content appears.

Formal criteria for that judgement were set out early and still structure practice. Cooper and Poinar published an influential statement of authentication standards in Science in 2000, arguing that work on old material required physically isolated facilities, control amplifications, appropriate molecular behaviour such as shorter products succeeding where longer ones failed, reproducibility, and independent replication, along with careful attention to whether results made phylogenetic sense.

Computation completes the separation that cleaning begins. Damage patterns are quantified across millions of fragments, endogenous proportions are estimated honestly rather than rounded upward, and statistical models test whether the surviving data fit the claimed history better than they fit a contamination scenario. A laboratory therefore ends where the overall argument began. Trust is not declared at the bench or conferred by a single impressive sequence. It accumulates across substrate choice, excavation care, room discipline, blank controls, replication, damage measurement, and statistical fit, until the remaining explanation for the pattern in the data is the survival of genuine old fragments, measured 30 to 70 base pairs at a time, inside material that outlasted the organism it once described.

Extraction: Freeing Short Molecules From Bone Powder

Extraction begins before any molecule reaches a tube, because the choice of material decides how much endogenous material survives to be freed at all. Dense cortical bone, the thick petrous portion of the temporal bone around the inner ear, and the cementum layer of a tooth root protect molecules differently, and laboratories weigh that protection against the destruction sampling causes. The outer surface of the chosen piece is removed or treated to strip away material deposited by soil and handling, and the remaining interior is reduced to a fine powder. Grinding multiplies the surface area exposed to the extraction chemistry, so reagents reach molecules trapped inside the mineral matrix rather than only those near a cut face. Every gram surrendered to powder is gone as anatomy, which is why photographs, sampling records and retained portions matter alongside yield.

Bone is a composite of collagen and hydroxyapatite, a calcium phosphate mineral that binds genetic material tightly, shielding it from some decay while holding it fast against release. Extraction is therefore a controlled dismantling. A chelating agent, most often EDTA, seizes calcium ions and slowly dissolves the mineral phase over hours or days, while proteinase K digests the protein scaffold that remains. As the matrix loosens, molecules that survived burial in fragments commonly 30 to 70 base pairs long pass into solution alongside everything else the bone contained. That solution is chemically crowded. Humic acids from soil, tannins, calcium salts and degraded protein travel with the target molecules, and several companions inhibit the enzymes used later, so a method that releases plenty of material but carries inhibitors forward has solved only half the problem.

The method associated with Rohland and Hofreiter addressed that balance through silica based purification adapted to ancient material. Molecules bind to silica in the presence of high concentrations of chaotropic salts, which disrupt the water shell around both the silica surface and the phosphate backbone, allowing contact that plain aqueous conditions prevent. Alcohol based washes then carry salts and many inhibitors away while the bound fraction stays in place, and elution in a small volume of low salt buffer releases the purified material. Variations using guanidine based binding buffers and spin columns built for short fragments became standard across laboratories, because the approach recovered material that earlier phenol based methods lost or left too dirty to amplify.

A decisive refinement came from Dabney et al. 2013, who showed that the shortest molecules, the fraction carrying much of the authentic signal, slipped through conventional silica columns tuned for longer modern fragments. Their silica extraction for short fragments adjusted the binding buffer with isopropanol and tuned salt and pH conditions, using silica suspended in solution so that very short molecules bound instead of passing through in the flow. The gain mattered most in difficult material, where surviving molecules cluster at the bottom of the length range. A method blind to fragments under about 50 base pairs does not merely recover less from such a sample. It recovers a biased subset, skewed toward longer molecules more likely to include modern contamination, and that bias propagates into later inferences about ancestry and relatedness. Yield figures deserve suspicion for the same reason. A spectrophotometer cannot separate endogenous fragments from microbial material, so laboratories quantify by targeted amplification instead, and judge an extract by what it produces downstream.

Building a Sequencing Library From Damaged Ends

Why are ancient DNA fragments so short?

Ancient fragments are short because decay breaks the backbone long before burial conditions stabilize. Water drives hydrolysis that severs bonds between bases, microbes consume exposed material, and temperature accelerates each reaction. Crosslinks and missing bases accumulate alongside the breaks. What survives in bone is therefore a population of remnants, commonly 30 to 70 base pairs, rather than intact chromosomes.

A sequencer cannot read a raw extract. Instruments read molecules carrying known synthetic adapters at both ends, in known orientation, and a library is the converted population that meets those conditions. Library preparation rebuilds each surviving fragment into readable form, and the rebuilding must work on ends that decay has chewed unevenly. In ancient material one strand may overhang the other, bases may be missing, and chemical lesions may block the enzymes asked to repair the junction. A preparation method that assumes clean ends quietly discards exactly the molecules carrying the oldest signal, because tidiness in this material is a property of the intruder rather than of the specimen.

The double stranded method set out by Meyer and Kircher handles those ends in sequence. End repair enzymes polish the fragments, filling overhangs where the template allows and removing protruding ends that cannot be completed, so each molecule presents blunt ends. Adapters are ligated onto those ends, and a polymerase fills the remaining nicks at the adapter junction, producing complete double stranded molecules that amplification can copy faithfully. Each step loses material, since enzymes fail on some lesions and purification retains only a fraction of what entered. The method nevertheless transformed practice when paired with high throughput sequencing, because it converted enough of a precious extract into a stable library that could be sequenced, stored and returned to later without resampling the specimen.

Indexing is the quiet step that makes large projects possible. During amplification each library receives a short synthetic barcode unique to that sample, and often a second tag on the other adapter. Pooled libraries from dozens of individuals then travel through one sequencing run together and are separated afterward by software reading the tags. Dual indexing also guards against misassignment, because a molecule carrying an unexpected tag combination can be discarded rather than trusted. The single stranded method published by Gansauge and Meyer in 2013 recovered much of what the double stranded route lost. By denaturing the extract and building the library one strand at a time, that approach recovers molecules with a broken strand and molecules too short to survive blunt end repair. Restraint governs the closing amplification either way. Extra cycles multiply duplicates and contaminants, so laboratories choose the smallest cycle number that yields enough, knowing that library complexity, the count of distinct original molecules, sets the ceiling on everything that follows.

Adapters themselves are engineered rather than generic. Their sequences must prime reliably, carry the binding site the flow cell requires, and avoid forming dimers with each other, because two adapters ligated without an insert between them create a short, efficiently amplified product that wastes reads and can dominate a weak library. Size selection after ligation removes much of that dimer fraction, and quantification against the adapter sequence measures only molecules that received adapters at both ends, which is the population that can actually be sequenced. A library can therefore look abundant by crude measurement and still be nearly empty of usable material.

Repair, Partial Repair and Leaving Damage in Place as Evidence

Damage is not one thing, and treating it as one thing leads to bad choices at the bench. Strand breaks end a molecule outright. Missing bases, called abasic sites, stall a polymerase that reaches them. Crosslinks tie strands together and resist denaturation, while oxidative lesions alter individual bases so they pair incorrectly. The most studied family, deamination, changes the identity of a base itself. Briggs et al. 2007 characterized that pattern systematically, showing that cytosine converted to uracil through deamination concentrates near fragment ends, where single stranded overhangs expose the base to water. During sequencing the lesion is read as a cytosine to thymine change at the start of a read, with the mirror change at the other end. That signature is both a defect to manage and, as a later section explains, a certificate of age to exploit.

Full enzymatic repair attacks deamination directly. Uracil DNA glycosylase, usually paired with an endonuclease in a commercial repair mix, recognizes uracil in the strand, excises it and cleaves the backbone there, so the damaged position is removed rather than copied. Libraries treated this way, often described as UDG treated, produce cleaner sequence. Spurious changes that could be mistaken for genuine variants largely disappear, genotype calls grow more reliable, and comparisons against reference genomes carry less systematic error. For questions turning on fine differences between populations, where a small excess of apparent derived alleles in one sample could mimic migration or selection, that cleanliness is worth a great deal, and many large studies build their main libraries with full repair as the default.

The price of full repair is the destruction of evidence. Once uracils are excised, the damage pattern distinguishing an ancient molecule from a modern intruder is gone from that library, and no later analysis can recover it. Repair also shortens molecules further, because cleavage at a lesion near an end trims the fragment. A laboratory that repairs everything has traded its best authentication signal for accuracy, and must authenticate by other means, through a separate untreated library, through replication, or through consistency across many individuals from the same context. The decision is irreversible for that aliquot, which is why extracts are split before treatment whenever quantity allows, preserving an untreated portion against questions nobody has thought to ask yet.

Partial repair offers a negotiated settlement. In the partial approach, sometimes called partial UDG, reaction conditions are adjusted so uracils in the interior of a fragment are removed while a share at the very ends survives. The interior of each read then behaves like repaired sequence, suitable for variant calling, while the terminal positions retain the elevated cytosine to thymine rate marking genuine age. One library serves both purposes, although the compromise is real. Terminal bases must still be masked or downweighted in sensitive analyses, and the retained damage rate is lower than in an untreated library, so authentication thresholds calibrated on untreated material do not transfer directly. Leaving damage entirely in place remains right for other questions. A study of decay itself needs the lesions as primary data, and mitochondrial work has proceeded untreated in many projects because the target exists in many copies per cell, letting depth overcome accuracy losses at individual positions. Repair is therefore a decision about which truth a library should tell, the truth of the sequence or the truth of its survival, and careful projects arrange to hear both.

Sequencing Millions of Fragments That Mostly Are Not Human

A finished library looks deceptively uniform, a clear solution of adapter carrying molecules within a narrow size range, but its contents are a census of the burial environment alongside a census of the individual sampled. Bone buried in soil is colonized after death by bacteria and fungi feeding on its nutrients, and their genetic material enters the same pores and cracks that shelter the host molecules. During life the body carried its own microbiome as well. Soil, water and the hands of everyone who handled the specimen add further contributions. Extraction and library preparation do not distinguish among these sources, because the chemistry binds molecules by size and structure rather than by origin. Sequencing therefore reads whatever was loaded, in proportion to whatever the library contains.

That proportion is the central economic fact of the field. In a poorly preserved sample the endogenous human share can be below one percent of the library, the remainder given over to microbial and environmental material, and even well preserved specimens may surrender half or more of their reads to nonhuman sources. Values vary enormously by site, tissue and burial history, which is why pilot sequencing precedes any large commitment. A shallow first run on a new library, sometimes called a shotgun screen, measures the endogenous share, the fragment length distribution, the damage rate and the duplicate rate before expensive deep sequencing is approved. A library that fails the screen is not sequenced further. The specimen may be resampled at a better element, the extract rebuilt by a different method, or the individual set aside, and each outcome is cheaper than filling an instrument run with bacterial sequence nobody needs.

The instruments explain why such screening is routine rather than wasteful. High throughput sequencers read in a massively parallel fashion, generating millions to billions of short reads in one run by copying clusters of identical molecules on a flow cell and recording the base added at each cycle through fluorescent signals. Read lengths in ancient work stay short, often 50 to 100 cycles, because the template molecules are short and reading past the end of a fragment only sequences the adapter. Paired end runs read a fragment from both directions, and for the shortest molecules the two reads overlap and merge into a single higher confidence sequence. Output arrives as enormous files of bases and quality scores, and from that moment the work belongs to computation rather than to the bench.

Depth is planned against a target outsiders often misunderstand. Coverage describes how many independent reads, on average, span each position in the genome, and ancient projects frequently work at low coverage, sometimes well under one fold, because population level questions can be answered statistically from many individuals each sequenced shallowly. Genotype likelihood and imputation methods extract information from sparse reads by modelling uncertainty instead of pretending each position was observed cleanly. Deep sequencing of one individual serves different questions, about heterozygosity, runs of homozygosity or a particular locus, and costs proportionally more of a shared run. The budget question is always how much information a fixed pool of reads buys when spread thin across a cemetery or concentrated on one skeleton. Before mapping even begins, adapters are trimmed, overlapping pairs merged, very short reads discarded because they align ambiguously, and amplification duplicates marked, so that multiple copies of one original molecule are never mistaken for independent observations.

Capture and Enrichment: Fishing for the Fragments That Matter

Shotgun sequencing reads a library in proportion to its contents, which is honest and wasteful in equal measure when the endogenous share is small. Capture changes the proportions before sequencing by selecting, from the finished library, only those molecules that hybridize to a chosen set of target sequences. The targets are supplied as baits, synthetic molecules whose sequences match the regions a project cares about, and the library is incubated with them under conditions letting matching fragments bind while nonmatching material washes away. The bound fraction is recovered, amplified modestly and sequenced. A library in which human fragments formed a small minority can emerge from capture with those fragments as the overwhelming majority of what the instrument reads.

Two bait technologies have carried most of the work. Solution capture uses biotin labelled baits mixed freely with the library, so hybridization happens quickly throughout the liquid, after which streptavidin coated beads pull the biotin bearing complexes, baits and captured fragments together, out of the wash. Array capture instead fixes baits to a solid surface and passes the library across it. Solution methods prevailed for ancient material largely on practical grounds. They scale to hundreds of samples, tolerate short fragment lengths, and can be repeated on the same library when a first pass underdelivers. Both approaches share a bias analysts must remember. Hybridization favours molecules matching the bait closely and molecules of moderate length, so capture data can underrepresent divergent sequences and the very shortest fragments, an ascertainment effect that matters when captured data are compared with shotgun data or with populations distant from the reference used to design the baits.

Target choice defines what a captured dataset can ever say. Early ancient work captured the mitochondrial genome alone, because its many copies per cell made it recoverable from poor material and because a complete mitochondrial sequence answers real questions about maternal lineage. Nuclear capture then expanded the ambition. The design that reshaped population history was a panel targeting a fixed set of single nucleotide positions chosen for ancestry analysis. Fu et al. developed capture approaches for ancient material that made targeted nuclear recovery practical at scale, and the panel known as the 1240k, built around roughly 1.24 million positions spanning the autosomes, the sex chromosomes and the mitochondrial genome, became a shared standard. Haak et al. 2015 applied genome wide capture of this kind to steppe and European material, and Mathieson et al. 2015 used the same style of panel across a large transect of ancient Europeans.

That shared target set is the deepest consequence of the design. Population genetics advances by comparison, and comparison requires commensurable data. A laboratory working on a cemetery of modest preservation can produce genotypes at the same positions as a flagship project working on exceptional material, and the two datasets join without either being resequenced. Migration and admixture are population questions that small samples answer weakly however deeply each genome is sequenced. Capture is not a rescue for a failed library, however. A molecule absent from the library cannot be captured, contaminated libraries yield contaminated captures because baits select by sequence match alone, and strong capture can distort allele balance when one allele matches a bait better than the other. Enrichment is a lens with known aberrations, powerful exactly because those aberrations are understood.

Authentication Through Damage Patterns

Every claim built on ancient molecules rests on a prior claim, that the sequences reported come from the specimen named and not from the laboratory, the excavator or the soil. Authentication is the discipline of proving that prior claim, and its standards were set early because the young field embarrassed itself without them. Cooper and Poinar, writing in Science 2000, laid down criteria for ancient work that still structure practice, demanding physically isolated facilities, regular monitoring for contamination, independent replication, appropriate controls, and molecular behaviour consistent with great age, including the inverse relationship expected between fragment length and amplification success. Their list was not bureaucratic caution. Early reports of genetic material from amber inclusions and from material of dinosaur age had collapsed under scrutiny, and the criteria drew a line between results a sceptic could test and results resting on trust.

The strongest everyday evidence of age is the damage pattern itself. Briggs et al. 2007 established that deamination leaves a predictable, quantifiable signature, an elevation of cytosine to thymine changes at the first positions of a read and the complementary change at its end, decaying inward as distance from the fragment end grows. Modern contaminant molecules carry no such signature, because their cytosines have not sat in single stranded overhangs in wet soil for millennia. Software built to measure the pattern, most prominently the mapDamage approach, plots substitution rates by position for every library and reports the damage fraction at the terminal base. A strong damage curve in an untreated or partially repaired library is evidence no procedural promise can match, because it is a property of the molecules rather than of the laboratory notebook.

Length tells a parallel story. Authentic molecule populations from old bone concentrate at short lengths, commonly 30 to 70 base pairs, with a distribution falling away toward longer fragments. A library dominated by long fragments behaves like modern material whatever its paperwork says, because intact long molecules survive only where decay was arrested early and completely, a claim requiring evidence of its own. Analysts therefore read the length histogram alongside the damage plot, and together the two form a fingerprint. Damage without shortness, or shortness without damage, weakens the case, while both together, replicated across independent extracts from the same specimen, make contamination an increasingly contorted explanation. Mitochondrial consensus checks add a third line, testing whether mitochondrial sequences in a library resolve to one coherent lineage or smear across several, as a mixture of modern handlers would produce.

Replication remains the criterion that settles disputes when fingerprints argue. Cooper and Poinar required key results to be reproduced, ideally in a second laboratory from a separate sample of the same specimen, and the logic has not aged. A contaminant peculiar to one building, one reagent batch or one technician cannot easily reproduce itself in another city under another protocol, while a genuine endogenous sequence reproduces wherever adequate material survives. Authentication also yields numbers rather than verdicts alone. Contamination estimates from mitochondrial mixtures, from sex chromosome ratios in males and from apparent heterozygosity on haploid chromosomes state how much of a library is foreign, so a library with a small measured fraction can still serve many analyses once that fraction is reported and modelled. A library whose authenticity rests on assertion alone can serve none.

Contamination: Every Route by Which Modern DNA Enters

Contamination is best understood as a set of routes rather than a single accident, because each route enters at a different point in the life of a sample and demands a different defence. The earliest route is the burial environment itself. Groundwater moving through bone carries microbial material deep into its pores, and soil organisms colonize the specimen for as long as it lies underground. Handling opens the human routes. Excavators, curators and students have touched most museum specimens with bare hands across decades of study, shedding skin cells into cracks, and specimens were historically washed, glued and varnished, each treatment adding human material and sometimes driving it deeper. Sampling design answers that history by discarding or aggressively cleaning the outer millimetres and taking material from protected interiors.

The laboratory is the second great source, paradoxically because it is full of exactly the material being sought. Modern human material arrives on the skin and clothing of staff, in dust, on inadequately cleaned equipment, and above all in amplified products from previous experiments. A completed polymerase chain reaction holds billions of copies of a target sequence in a small tube, and aerosol droplets from opening such a tube can seed a room or a pipette with more copies than an ancient extract contains in total. Ancient facilities therefore enforce a one way geography. Extraction and library preparation happen in a dedicated clean laboratory where no amplified product and no modern reference material is admitted. Staff dress in protective clothing and masks before entering, surfaces are treated with bleach and ultraviolet light, and work flows physically forward toward the ordinary laboratories where amplification and sequencing occur, never back.

Reagents and consumables form a third route that geography cannot close. Commercial kits, enzymes, water and plasticware can carry trace human or microbial material from manufacture, and because ancient extracts contain so few target molecules, a trace invisible in a modern experiment can rival the genuine signal. Blank extractions exist to expose exactly this. A blank is a sample with no bone powder, carried through every extraction step alongside the real specimens, then built into a library and sequenced like any other. Whatever appears in the blank arrived from reagents, consumables or the laboratory environment during that batch, and its composition names the culprit class. Negative controls extend the same logic through library preparation and amplification, where a reaction containing no template should produce nothing, and any product it does produce undermines trust in its batchmates.

Cross contamination between samples is the fourth route, and it grows more dangerous as projects scale. Samples are drilled in sequence with shared equipment, extracts sit in adjacent wells, and libraries are pooled before sequencing, so a droplet, a misread label or a reused tip can move material from an abundant sample into a sparse one. At the bench the defences are discipline, one open tube at a time, blanks interleaved between specimens, and equipment decontaminated between drillings. Computational checks then audit what procedure attempted, flagging a library whose mitochondrial lineage matches the sample drilled immediately before it. The final route runs through analysis itself, where shared drives and merged datasets can mislabel as easily as a pipette can misdeliver, so provenance records travel with every file. A genome is only ever as trustworthy as the weakest of these defences, which is why contamination control is not housekeeping before the science but part of the science itself.

Mapping Fragments to a Reference Genome

A sequencer does not return a genome. A sequencer returns millions of short reads, most between 30 and 70 base pairs long, each read a fragment copied from a molecule that survived in bone, tooth, hair or sediment. Mapping is the step that places each fragment against a reference genome, a complete assembled sequence from a living person or from a closely related species, and asks where the fragment fits best. Software aligns the letters of the fragment to the letters of the reference, allowing for mismatches caused by damage, by sequencing error and by genuine difference between the ancient individual and the reference. A fragment that aligns to one position only carries more weight than a fragment that aligns equally well to many positions, because repeated regions of the genome produce ambiguity that no amount of additional sequencing can fully resolve.

The reference shapes the result in ways that require constant attention. A fragment carrying the version of a letter found in the reference aligns more easily than a fragment carrying a different version, and that tendency, called reference bias, can understate divergence in divergent samples. The effect matters most when the ancient genome is distant from the reference, as with a Neanderthal mapped to a modern human reference or a Denisovan mapped to the same coordinates. Analysts reduce the distortion by realigning, by masking regions of low complexity and by checking whether patterns survive when a different reference is used. A result that appears only against one reference is treated as a property of the mapping rather than a property of the past population.

Short length imposes its own discipline. A read of 35 base pairs contains limited information, and a single mismatched letter represents a large share of its total evidence. Damage concentrates at the ends of molecules, where cytosine deamination converts one letter into another, so pipelines trim ends or downweight terminal positions before calling variants. Duplicate reads present a second problem. Amplification during library preparation copies the same original molecule many times, and counting every copy as independent evidence inflates confidence. Deduplication collapses reads with identical start and end positions into one observation, preserving the independence that statistical tests assume.

Mapping also supplies quality measures that govern every later claim. The share of reads that map at all separates a well preserved sample from a sample dominated by soil bacteria. Average coverage, expressed as the number of independent fragments overlapping each position, determines whether a genotype can be called directly or must be inferred probabilistically. At low coverage, analysts often work with pseudohaploid calls, sampling one read at each known variable position rather than pretending that both copies of a letter are known. The method loses information about heterozygosity but avoids systematic error, and large population studies built on pseudohaploid data have proved robust when sample sizes are large.

Alignment is therefore not a clerical step between sequencing and interpretation. Alignment decides which molecules count, which positions are trustworthy and which differences are real. A laboratory can generate impeccable sequence from a contaminated extract and map that sequence beautifully, and the map will still describe the wrong organism. Conversely, a modest number of fragments, mapped with strict uniqueness thresholds and damage aware scoring, can support a durable placement of an individual in time, place and population. Every downstream section of a study inherits the mapping choices.

Mitochondrial DNA, the Y Chromosome and the Whole Genome

Three kinds of sequence circulate in ancient remains, and each answers a different class of question because each follows a different rule of inheritance. Mitochondrial sequence passes from a mother to all children, exists in many copies per cell and survives more often than nuclear sequence in degraded material. The high copy number explains the early history of the field, when mitochondrial fragments were the only recoverable target in most specimens. A mitochondrial lineage places an individual within a maternal genealogy, identifies species when morphology fails and provides a first test of preservation before expensive nuclear work begins. The limitation is scale. The mitochondrial genome is small, represents one line of descent among thousands and can disagree with the history of the rest of the genome when populations mixed after separating.

The Y chromosome occupies the opposite corner. Passed from father to son with little recombination across most of its length, the Y records a single paternal line with similar clarity and similar narrowness. In male burials, Y lineages reveal patrilineal structure, the arrival of male mediated migration and the expansion of particular paternal lines during periods such as the Bronze Age. Skoglund and colleagues formalized a related use of sex chromosomes for sex determination, comparing the ratio of fragments aligning to the X chromosome against fragments aligning to the Y chromosome. A sample with Y aligned fragments at a predictable fraction is male, a sample with almost none is female, and the test works at coverage too low for reliable genotyping across the autosomes. The method reshaped burial interpretation because osteological sex estimation fails for juveniles and for fragmentary remains.

Whole genome data from the autosomes, the twenty two numbered chromosome pairs, supports the broadest inference. Autosomal sequence samples ancestry from all recent ancestors rather than one line, allows estimation of relatedness between individuals and permits comparison of allele frequencies across populations. The cost is preservation and expense. Autosomal coverage adequate for diploid genotyping demands exceptional material, such as the petrous portion of the temporal bone or well preserved teeth, and many large studies settle for targeted capture of known variable positions instead of unrestricted sequencing. Capture reagents enrich libraries for selected loci, raising usable coverage at those positions while leaving the rest of the genome sparse.

Choice among the three systems is a choice about the question. A study asking whether two cemeteries share maternal continuity can answer with mitochondrial data alone, provided sample sizes are adequate. A study asking whether a migration was sex biased needs both uniparental systems and autosomal data together, because a pattern visible in Y lineages and absent in mitochondrial lineages carries a different meaning from a pattern visible in both. Mismatch between systems is information rather than contradiction. Populations do not move as uniform blocks, and the separate histories recorded in a mother line, a father line and the autosomal average expose structure that any single system would flatten into a misleading average. Interpretation also requires attention to time depth, because a uniparental lineage records the survival of one line among many that left no descendants in that system, while autosomal ancestry averages across lines that persisted through different routes, places and generations over many centuries of mixing and separation thereafter.

Sex, Kinship and Family Reconstruction From Shared Segments

Sex determination from sequence data changed burial archaeology because sex shapes grave goods, body position, diet and status in many societies, and misassigned sex distorts every social reading built upon a cemetery. The fragment ratio method associated with Skoglund and colleagues gave laboratories a molecular test independent of skeletal completeness, comparing fragments aligning to the X chromosome against fragments aligning to the Y chromosome. Juvenile burials and fragmentary remains became legible in a new way, and cases accumulated where molecular sex contradicted assumptions drawn from objects placed in a grave. A weapon beside a skeleton no longer settled the question, and a bead necklace no longer settled the opposite question.

Kinship analysis begins from a simpler observation. Close relatives share longer and more numerous segments of sequence inherited from a common ancestor, and the total amount shared falls in predictable bands for parent and offspring, for siblings and for more distant relationships. Even at low coverage, statistical methods compare allele sharing between two individuals against expectations for unrelated pairs from the same population and assign a degree of relatedness with stated uncertainty. Parent and offspring pairs are the most secure. Sibling pairs share more variable amounts because recombination shuffles contributions from each parent, and distant relationships blur into population level similarity in small communities where unrelated individuals already share elevated background ancestry.

Applied to cemeteries, kinship reconstruction turns a collection of graves into a social document. Studies of Neolithic and Bronze Age burial groups in Europe identified multi generation pedigrees arranged in space, with related individuals clustered and unrelated individuals, often women, buried within household groups whose profiles pointed to distant birthplaces. Such patterns support accounts of patrilocal residence, in which sons remained near birthplaces and partners arrived from elsewhere, although other cemeteries show no spatial kin structure at all. Adoption, fostering and remarriage create families that genetics alone cannot see, and low coverage, contamination and reference population choice widen every estimate, so the strongest reconstructions combine relatedness with isotope evidence, grave architecture and chronology.

Can DNA be recovered without a skeleton?

Yes. Sediment from cave floors and dental calculus built up on teeth both preserve recoverable genetic material after bone has vanished, decayed or never been deposited in a sampled layer, and each reservoir extends recovery beyond the limits set by surviving skeletons in a burial or occupation deposit alone.

Sediment work depends on the observation that organic material leaches into and binds with mineral particles in cave floors, lake beds and occupation surfaces. Fragments persist in layers that contain no visible bone, and samples through a stratigraphic column can track which hominin groups occupied a site across time. Denisova Cave provided the defining demonstration, with sediment profiles revealing alternating Neanderthal and Denisovan presence in layers lacking diagnostic fossils. The approach demands strict controls, because material moves between layers through water, burrowing animals and trampling.

Dental calculus offers a more personal reservoir. Mineralized plaque accumulates during life and traps food particles, oral bacteria and host cells in a hard matrix that resists decay after death. Calculus survives on teeth when softer tissues have disappeared, preserves a record of diet and oral microbiome composition and yields host fragments sufficient for population assignment in favorable cases. Because calculus forms during life, associated microbial profiles reflect a living mouth rather than a burial environment, a distinction that matters for pathogen and diet studies alike. Together, sediment and calculus widen the archive considerably.

Ancestry and Migration: Admixture Models and Their Limits

Ancestry in population genetics has a technical meaning narrower than everyday usage. Ancestry describes the sources from which a genome descends, modeled as mixtures of earlier populations that themselves carried distinctive allele frequency patterns. Admixture models formalize that description. A target population is represented as a combination of proposed source populations in estimated proportions, and the adequacy of the representation is tested against observed patterns of allele sharing. Methods based on allele frequency correlations ask whether a simple mixture accounts for the relationships among groups, or whether additional sources, later gene flow or a different history altogether fits the pattern better. A model that passes formal tests has survived an attempt at rejection. A model that fails points toward missing sources or a more complex sequence of events.

Lazaridis and colleagues in 2014 gave European prehistory a durable form of this reasoning with a three ancestral populations model for early Europeans. Ancient hunter gatherers from western Europe, early farmers whose ancestry traced to the Near East and a third component related to ancient populations of northern Eurasia together accounted for the ancestry of most later European groups in differing proportions. The model succeeded because named ancient genomes, rather than modern proxies alone, anchored each component. Later work refined, subdivided and in places replaced components of the scheme, which is the expected fate of a good model. The value lay in testability. Any new genome from early Europe could be placed against the three source framework and either fit, extend the framework or break the framework in an informative direction.

Haak and colleagues in Nature in 2015 supplied the migration test case that reshaped debate about the spread of Indo European languages into Europe. Genomes from Corded Ware associated individuals in Germany carried a large share of ancestry related to Yamnaya herders of the Pontic steppe, dated to a movement around five thousand years ago, and the size of the component implied substantial population movement rather than cultural diffusion alone. Earlier farmers of central Europe carried little of that component, and later populations carried intermediate amounts, consistent with mixing after arrival. The result did not settle the language question by itself, because languages leave no molecules, but the scale of movement removed a class of explanations that required steppe influence without steppe people.

Limits follow from construction. A source population in a model is a sample of genomes given a label, not a bounded historical people with a single language, territory and self understanding. Labels such as farmer, hunter gatherer and steppe herder compress internal diversity and can harden into ethnic claims that the allele data never supported. Equivalent models also exist. Different combinations of sources sometimes fit the same target nearly equally well, and choosing among equivalent fits requires archaeological and chronological evidence outside genetics. Admixture dates estimated from the decay of linkage patterns carry wide intervals, and a date compatible with an archaeological horizon does not prove that a particular cultural change caused a mixing event.

Geography exposes a further limit. Sampling follows preservation, excavation history and research funding, so maps of ancient ancestry contain blank regions that reflect absence of data rather than absence of people. A migration route drawn through sampled points may bend around an unsampled region that held the actual corridor. Molecular clocks calibrated against dated genomes also serve far deeper questions, including when animals first evolved and why genetic estimates of that origin predate the fossil record.

Species Identification, Neanderthals and Denisovans

Bones often arrive without diagnostic shape. A fragment from a cave layer may be a splinter too small for anatomical placement, a juvenile element lacking adult features or a piece altered by burning, gnawing or sediment pressure. Genetic identification answers a taxonomic question before any population question can begin. Mitochondrial sequence usually suffices for species level placement because mitochondrial differences between closely related hominins exceed variation within a lineage at many positions, and short fragments carry enough of those differences to separate human, Neanderthal and Denisovan material from the far larger background of animal bone in most cave deposits. Collagen peptide fingerprinting offers a complementary screen for large assemblages, sorting thousands of fragments cheaply so that sequencing effort concentrates on pieces likely to be hominin.

Neanderthals became the first extinct hominin group defined at genome scale. Green and colleagues in Science in 2010 published a draft Neanderthal genome assembled from bones from Vindija Cave in Croatia, and comparison with genomes from living populations produced the estimate that genomes outside Africa carry Neanderthal ancestry between 1 and 4 percent. The figure rested on patterns of allele sharing that exceeded sharing with genomes from Africa, consistent with interbreeding after the ancestors of living non African populations left Africa and encountered Neanderthals in western Asia. Later high coverage genomes refined which segments survived, showed that some regions of living genomes systematically lack Neanderthal ancestry and identified individual inherited segments associated with traits involving skin, hair and immune response. The draft genome mattered because scale changed the question from whether any mixing occurred to how much, where and with what lasting consequences.

Denisovans entered knowledge through sequence before anatomy. Krause and colleagues in 2010 reported mitochondrial sequence from a finger bone from Denisova Cave in Siberia that diverged from both Neanderthal and modern human mitochondrial lineages, signaling an unrecognized group. Reich and colleagues in 2010 extended the result with nuclear sequence, and Meyer and colleagues in Science in 2012 produced a high coverage Denisovan genome from the same finger bone, enabling detailed comparison despite a fossil record consisting of a few teeth, small fragments and later a jawbone from the Tibetan Plateau identified through protein evidence. Populations in Oceania carry the highest Denisovan ancestry proportions found in living groups, with additional contributions detected across Asia and evidence for more than one Denisovan related source mixing with separate human populations.

The Denisovan case rewrote expectations about discovery. A group with a wide geographic reach and measurable legacy in living genomes remained invisible to a century of fossil hunting because diagnostic bones were rare, fragmentary or unrecognized in existing collections. At the same time, a genome without a face imposes restraint. Claims about Denisovan appearance, behavior or language drawn from sequence alone exceed the evidence, and responsible accounts separate the secure findings, relatedness, divergence and admixture, from speculation about a population known mainly through molecules and a handful of mineralized tissues.

Identification work also returns value to fauna and environment. The same extracts that yield a rare hominin fragment contain animal sequence that reconstructs past ecosystems, tracks species turnover across climate shifts and identifies prey, predators and commensals at a site. A cave deposit becomes a biological archive in bulk, and hominin findings gain context from the community of species recorded alongside them.

Pathogen Genomes Recovered From Teeth

Teeth preserve more than the genome of the person who carried them. Bloodborne bacteria and viruses circulating during life can leave fragments trapped in dental pulp, and microbes of the mouth accumulate in calculus, so a single tooth holds both a host archive and a microbial archive with different histories. Recovery of pathogen genomes from teeth and from bone near blood rich tissue allows direct study of infections that shaped mortality long before written medical records, and allows comparison of ancient strains with living descendants across known intervals of time. The distinction between infection present at death and contamination acquired in soil is central, and authentication leans on damage patterns, fragment length and anatomical plausibility together.

Bos and colleagues in Nature in 2011 reconstructed a Yersinia pestis genome from Black Death burials in London, drawing material from teeth of individuals buried in a cemetery associated with the epidemic that swept Europe from 1347 to 1351 during the first wave. The reconstructed genome showed that the medieval strain stood near the ancestry of later plague lineages, and the result settled a long dispute about the identity of the causative organism by direct molecular evidence from dated burials. Earlier attempts based on short targeted fragments had produced contested claims, and genome scale recovery with strict contamination controls carried a different evidentiary weight. Plague work since that reconstruction has traced earlier and later outbreaks, revealing a branching history of the bacterium across Eurasia rather than a single repeated introduction from one source.

Other pathogens leave different signatures. Mycobacterium tuberculosis complex bacteria produce skeletal lesions in a minority of cases, yet molecular recovery succeeds in some lesion free material, complicating any simple link between visible pathology and infection. Hepatitis B virus fragments recovered from remains thousands of years old extended the known history of that virus and revealed extinct lineages with no living descendants. Oral microbes in calculus, while seldom lethal, record shifts in diet and in the ecology of the mouth associated with farming transitions and with industrial food processing. Each case turns on the same evidentiary chain. The organism must fit the tissue, the damage pattern must fit age and the phylogenetic placement must make sense against dated samples from other sites.

Pathogen genomes also calibrate evolutionary rates in a way that modern sampling alone cannot. A genome from a securely dated burial anchors a point in time, and a series of anchored genomes across centuries allows estimation of how fast substitutions accumulated. Rates estimated without ancient anchors depend on assumptions about generation times and past population sizes, and anchored estimates have revised timelines for several bacterial and viral histories. Revised timelines change historical inference in turn, separating an ancient origin with slow spread from a recent origin with fast spread, two accounts that modern genomes by themselves may fit almost equally well.

Restraint governs interpretation of cause of death. Presence of a pathogen genome in pulp establishes infection or at least bloodstream presence near death, not a certified cause, and coinfection, malnutrition and violence leave overlapping traces in burial populations. Epidemic cemeteries strengthen inference through numbers, context and historical association, while an isolated positive skeleton supports a narrower claim about circulation of a strain in a place and period. The wider consequences of the epidemics those pathogen genomes identify are traced in the account of how pandemics changed history.

The History of a Field That Nearly Failed on Contamination

The field began with a sound result and a fragile method. Higuchi and colleagues in 1984 recovered short mitochondrial fragments from a museum specimen of the quagga, an extinct relative of the zebra, demonstrating that genetic material could survive death and be read from preserved tissue. Paabo in 1985 extended the attempt to Egyptian mummy material, reporting cloned fragments from a mummy thousands of years old. The mummy work later drew scrutiny over contamination and damage, and the episode taught an early lesson that survival in a museum skin, dried and stable, differs from survival in most buried remains. Enthusiasm outran controls. Polymerase chain reaction, arriving soon after, amplified trace material with extraordinary sensitivity, and that sensitivity amplified modern contamination just as efficiently as ancient molecules.

The 1990s produced the collapse that nearly ended credibility. Claims appeared for genetic material from insects preserved in amber tens of millions of years old, including the report by Woodward and colleagues in 1994, and for sequences attributed to dinosaur bone. Independent laboratories could not replicate the amber and dinosaur results, sequences matched modern organisms in suspicious ways and geological age alone made survival of readable fragments implausible under known rates of chemical decay. Journals and funders absorbed the damage. A field that promised resurrection of extinct genomes became associated in many scientific quarters with artifact, wishful thinking and insufficient controls, and careful workers spent years separating a legitimate core from a literature that included claims no authentication standard could support.

Recovery came through standards rather than a single breakthrough. Laboratories adopted dedicated clean facilities physically separated from amplified product, protective clothing, bleach and ultraviolet treatment of surfaces and reagents, extraction blanks carried through every step and independent replication in a second laboratory for extraordinary claims. Analytical criteria hardened alongside physical controls. Authentic ancient fragments should be short, should show characteristic damage near molecule ends, should produce consistent results across separate extracts and should yield phylogenetic placements that fit the specimen rather than a laboratory worker, a reagent supplier or a previous sample. Cloning gave way to high throughput sequencing in part because sequencing revealed the full mixture in an extract, including the bacterial majority and any modern human minority, instead of hiding the mixture behind selected clones.

Contamination never disappeared as a problem, and mature practice treats contamination as a measurable quantity rather than a moral failing. Estimates based on mitochondrial consensus mismatches, on X chromosome patterns in males and on damage restricted subsets allow analysts to state a contamination rate for each sample and to test whether conclusions survive after contaminated reads are modeled or removed. Graded usability replaced the older binary of pure and ruined, and reporting standards expect the grade to be stated.

The near failure left a durable culture. Major claims in ancient genomics attract immediate attempts at replication, methods sections disclose controls in detail and reviewers ask about blanks, replication and damage before asking about historical narrative. That skepticism is protective. Each genome that reshaped understanding of Neanderthals, Denisovans, steppe migration or plague survived the standards built from the amber and dinosaur episode, and the contrast between a replicated genome and an unreplicated claim remains the clearest teaching example the field possesses. Trust in ancient genomics rests not on the age of a specimen or the prestige of a laboratory, but on a chain of controls that any skeptical reader can inspect, repeat and challenge.

Paabo, the Neanderthal Genome and the Standards That Followed

The field acquired a durable standard when isolated successes gave way to a repeatable system, and no career illustrates that transition more clearly than the work led by Svante Paabo. Trained as a molecular biologist, Paabo pursued genetic material from extinct organisms through a period when many laboratories regarded such claims with justified suspicion. Early extracts produced sequences, yet contamination from handlers, reagents, and museum environments could produce equally readable results. The decisive change was methodological rather than rhetorical. Extraction moved into dedicated clean facilities, protective clothing and ultraviolet treatment became routine, negative controls accompanied every batch, and independent replication ceased to be an optional courtesy. A result earned confidence when another laboratory, working from a separate sample of the same specimen, recovered a compatible pattern.

An early landmark came from Matthias Krings and colleagues in Cell in 1997. The team recovered a mitochondrial sequence from the original Neanderthal specimen found in the Neander Valley, and the sequence fell outside the range of variation observed in living humans. The finding did not reconstruct a genome, settle interbreeding, or describe Neanderthal biology in detail. Its importance lay in disciplined authentication. The authors reported controls, cloned products, and checks against contamination, and the work showed that a short endogenous fragment could survive in a fossil tens of thousands of years old. For a field troubled by spectacular and irreproducible claims, a modest result with transparent safeguards carried exceptional weight.

The scale changed with Green et al. in Science in 2010. Richard Green and a large international team assembled a draft Neanderthal genome from bones recovered at Vindija Cave in Croatia, supplemented by material from other sites. Billions of short fragments had to be sorted from microbial sequences and modern contamination, then aligned against human and chimpanzee references. The resulting comparison supported a consequential estimate: Neanderthal ancestry accounts for between 1 and 4 percent of the genomes of many people whose ancestry lies outside Africa. That range, rather than a single exact figure, remains the careful form of the claim because coverage varied among fossils and statistical models made different assumptions about population structure.

Paabo received the Nobel Prize in Physiology or Medicine in 2022 for discoveries concerning the genomes of extinct hominins and human evolution. The award recognized more than one famous result. It marked the construction of an evidentiary culture in which extraction chemistry, sequencing technology, computational comparison, and skeptical review advanced together. The identification of the Denisovan lineage from a small finger bone from Denisova Cave belonged to the same transformation. A genome could reveal a population that the fossil record had not named, provided that contamination estimates, damage patterns, and population comparisons supported the inference.

Standards after the Neanderthal genome therefore extended beyond obtaining sequence data. Laboratories documented sampling location inside a bone, preserved extracts and libraries, reported contamination estimates, released raw reads, and distinguished observed bases from bases inferred through imputation. Journals and reviewers learned to ask how fragments were authenticated, how modern handlers were excluded as sources, and whether conclusions remained stable when damaged ends were trimmed. Success also raised expectations for older and poorer material. A published genome became the beginning of scrutiny rather than its end, and that expectation remains the strongest legacy of the Paabo era.

What a Single Genome Can and Cannot Establish

A single genome is a rich record from one life, and its limits begin with that simple fact. It can establish biological sex through the relative coverage of sex chromosomes, identify close biological relationships when comparable genomes exist, and reveal stretches of shared ancestry that point toward particular population connections. It can carry variants associated with pigmentation, metabolism, immune response, and other traits, although association in living populations does not guarantee the same effect in a different environment or genetic background. It can also contain pathogens or oral microbes in special contexts, and its damage profile can support the antiquity of the sampled material. Each of those findings concerns the sampled individual first. Broader claims require an additional step of reasoning.

That additional step is where overstatement often enters. One person sampled at one place and time cannot show how common a trait was across a population, which language a community spoke, how its members understood kinship, or why a migration occurred. Burial location does not turn an individual into a representative sample. A person may have travelled far from a birthplace, belonged to a minority lineage within a settlement, or left no descendants despite belonging to a large group. Conversely, absence of a lineage in one genome proves only absence in that person at detectable levels. It cannot erase a population, a contact event, or a cultural exchange that left no trace in that particular line of descent.

Comparison changes the strength of inference without removing uncertainty. When several genomes from a cemetery share segments inherited from a common ancestor, a kinship claim gains structure. When genomes across a region form a gradient over time, migration and mixture become more plausible than they would be from a single outlier. When one genome sits sharply apart from neighbouring samples, the careful conclusion may be mobility, uneven sampling, or an unsampled neighbouring population. The data support competing explanations until archaeology, isotope evidence, chronology, and additional genomes narrow the field. Genetic clustering is a statistical description of sampled material, not a direct portrait of social identity.

Trait interpretation demands the same restraint. A variant linked with lactase persistence, altitude response, or disease risk can suggest a biological capacity, yet capacity is not biography. Diet depends on available foods and cultural practice, disease depends on exposure, and survival depends on chance as well as inheritance. Polygenic scores trained on living populations transfer poorly across deep time because allele frequencies, environments, and linkage patterns differ. A responsible report names the variant, states the evidence behind its association, and separates that finding from speculation about appearance, behaviour, intelligence, or temperament. Claims moving from a handful of markers to a complete portrait of a person exceed the evidence.

Depth of coverage sets another boundary. A high coverage genome permits confident calls at most positions, detection of heterozygous sites, and more reliable estimates of runs of homozygosity. A low coverage genome may still place an individual within a broad population pattern, but many positions remain missing or are inferred statistically from reference panels. Such inference works best for populations well represented in those panels and less well for distinctive or poorly sampled groups. The distinction matters when conclusions concern rare variants or precise ancestry fractions. A single genome can open a question with unusual power, anchor a comparison, or falsify a simple story of isolation. It cannot carry the whole history of a people on its own.

The Measurement Problem: Reference Bias, Missing Data and False Relatives

Measurement begins before a sequencer produces its first read. Ancient molecules survive mainly as fragments 30 to 70 base pairs long, and many are shorter still or carry chemical lesions that block copying. Capture probes and mapping software must decide which fragments belong to the organism under study and which come from soil microbes, fungi, bacteria, or modern handlers. Every decision favours some kinds of evidence over others. Material that matches expectations is easier to recover, while divergent or heavily damaged material may be discarded during the very steps meant to reveal it. The resulting dataset is therefore a filtered remnant, not a neutral copy of a once complete genome.

Reference bias is the most systematic form of that filtering. Reads are commonly aligned to a reference genome drawn from a living human, and a fragment carrying the reference allele often maps more readily than a fragment carrying a different allele. The effect is small at any one position but accumulates across millions of positions. It can depress estimates of divergence, distort heterozygosity, and make an ancient individual appear closer to the reference population than the underlying evidence warrants. Capture design can add a second bias when probes are built around variants discovered in living groups. Populations distant from those groups then yield fewer captured sites, and analyses built on captured sites inherit that imbalance.

Missing data complicates every correction. Low coverage leaves positions unobserved, and damaged bases near fragment ends may be trimmed or statistically downweighted. Genotype likelihood methods handle uncertainty more honestly than forced calls, yet many downstream analyses still require a value at each site. Imputation fills gaps by borrowing patterns from reference panels, an approach that can be powerful for common variation and hazardous for rare or population specific variation. A filled value is a model output, not an observation. When reports fail to separate observed from imputed positions, later readers may treat a plausible reconstruction as though a sequencer had measured it directly.

False relatives arise when relatedness methods meet sparse and uneven data. Kinship estimation compares patterns of matching and mismatching alleles, but contamination can manufacture matches, low coverage can hide genuine ones, and shared population ancestry can resemble close family connection. Two unrelated people from a small or endogamous community may share more genetic material than a simple model expects, while genuine relatives sampled at different coverage levels may appear less connected. Laboratory error adds another route to error when samples are switched, indexed reads bleed between libraries, or extracts from neighbouring specimens mix. Careful laboratories use replicate libraries, contamination checks, and methods designed for low coverage data, and careful authors report degrees of relationship as estimates with uncertainty.

Chronology and preservation interact with all of these problems. Heat, moisture, acidity, and repeated freezing alter survival, so samples from one region may yield long fragments while samples of the same age from another region yield almost nothing. Comparing ancestry proportions across such uneven material can confuse preservation with history. A population may seem absent because its remains were sampled from teeth rather than dense petrous bone, or because its divergent reads mapped poorly. Strong analysis therefore reports fragment length, coverage, contamination, mapping choices, and sensitivity tests alongside the headline result. The history lies inside those technical details, and an estimate detached from them asks for more trust than measurement can supply.

The Strongest Critique: Authentication Can Become Circular

The strongest critique of the field does not deny that ancient molecules survive. It asks whether the tests used to certify survival can, under pressure to publish, begin to certify themselves. Alan Cooper and Hendrik Poinar gave that warning a durable title in Science in 2000: Ancient DNA: do it right or not at all. Their argument demanded physically isolated facilities, control extractions, cloning, independent replication, and biochemical evidence consistent with preserved material. The warning mattered because spectacular claims about very old material had repeatedly collapsed under contamination or irreproducibility.

Circularity enters through the criteria themselves. Short length and cytosine deamination near fragment ends count as signs of antiquity because ancient material displays those properties and modern material usually displays them less strongly. When a researcher first filters data for the expected damage signature and then presents that same signature as proof that the filtered data are ancient, the reasoning narrows into a loop. The test remains useful, but its force depends on unfiltered reporting, negative controls, contamination estimates from mitochondrial and nuclear data, and replication from a separate extract. Without those surrounding checks, a signature can become a label applied after selection rather than an independent test.

What does a damage pattern prove about a DNA fragment?

A damage pattern shows that a fragment behaves as postmortem material rather than fresh tissue. Cytosine deamination concentrated near molecule ends, short length, and strand patterns together support antiquity. That pattern does not identify the individual, fix a calendar date, or prove that every read in the extract shares the same origin.

Later skepticism about Egyptian mummy work exposed a related weakness. Reconstructions of the Tutankhamun family drew wide attention, yet critics questioned whether limited marker data from poorly preserved material in a hot environment could support the named parentage and medical conclusions attached to particular mummies. Identification depended on a chain of assumptions about which remains belonged to which historical figures, and genetic similarity was then used to reinforce relationships that the archaeological identifications had already suggested. The dispute did not prove every reported relationship false. It showed how quickly kinship diagrams can harden into biography when sample identity, preservation, and statistical support remain contested.

Heat makes the Egyptian case instructive beyond one royal family. Warm burial environments accelerate fragmentation and chemical damage, so yields may be low and modern contamination can form a larger share of recovered molecules. A genome wide result from such material deserves attention precisely because success is difficult, but difficulty raises rather than lowers the evidential bar. Single laboratory claims without released data sit in a weaker class. The distinction protects good work. It prevents the reputation earned by rigorously authenticated genomes from being transferred automatically to fragile results that share only a subject matter.

The thesis survives the critique in a narrowed and stronger form. Where separate laboratories recover compatible results, damage appears before selective filtering, contamination is quantified rather than asserted away, and population conclusions persist across methods, ancient sequence evidence retains exceptional force. Where a claim rests on one extract, expected signatures alone, or identifications imported from the conclusion under test, the proper verdict is suspension rather than belief. The critique therefore does not close the field. It defines the boundary between a measured molecule and a story told around a molecule, and it explains why method sections matter as much as headline findings. The disputed family reconstruction at the center of that skepticism is examined in the archive account of Tutankhamun, the boy king of Egypt.

Destructive Sampling, Descendant Communities and the Ethics of Taking Bone

Every ancient genome begins with loss. A drill removes powder from a tooth root or from the dense petrous portion of the temporal bone, and the removed material cannot be restored. Sampling may consume a unique fossil, cut through surfaces used by anatomists, or destroy contextual information that a future technique could have read from the intact specimen. The petrous bone yields comparatively rich material because its density protects molecules, and that scientific advantage creates an ethical concentration of harm. Curators are asked to sacrifice the very part of a skull that preserves both anatomy and molecular evidence, often before the probability of success can be known. A failed extraction leaves absence where a complete specimen stood.

Human remains add obligations that animal fossils and sediment samples do not carry. The dead person belonged to a family, a community, and often to a people whose descendants remain identifiable. Colonial collecting removed many remains from their places of origin under coercion or without consent, and museum possession does not settle moral permission for destructive research. Descendant communities may regard ancestors as relatives rather than specimens, may object to sampling on religious grounds, or may support research under conditions involving consultation, shared control, and return of results. Treating consultation as a public relations step after a sampling decision reverses the proper order. Permission, purpose, and limits belong at the design stage.

Alpaslan-Roodenberg and colleagues addressed that disorder in 2021 through five globally applicable guidelines for research on ancient human remains. The guidelines urged compliance with applicable laws and regulations, preparation of a detailed plan before sampling, efforts to minimise damage, attention to open and responsible data practice after publication, and meaningful engagement with stakeholders and descendant communities. Their importance lies less in novelty than in universality. Earlier ethical practice varied sharply by country, institution, and archaeological tradition, so a team could satisfy local procedure while ignoring communities with a direct stake in the remains. A shared floor does not remove disagreement, but it makes evasion harder.

Disagreement remains genuine because interests do not always align. A community may refuse sampling entirely, request reburial before analysis, restrict publication of sensitive findings, or ask that local researchers lead the project. A museum may hold legal title while lacking the trust of the community from which remains were taken. Scientists may argue that a genome could illuminate population history valued by descendants themselves, yet anticipated benefit does not override refusal. Data governance raises a second dispute. Open release supports replication and prevents private hoarding, but unrestricted genomic data can expose information about living relatives and can be reused for purposes a community never approved. Ethical practice must therefore manage access with the same care applied to bone.

Restraint can improve science rather than merely constrain it. Screening methods that estimate preservation before drilling, sampling from already broken surfaces, retaining part of every extract, imaging specimens before destruction, and coordinating multiple research questions around one sample all reduce waste. Collaboration can also improve interpretation because local knowledge may identify burial context, oral tradition, or historical disturbance that laboratory data alone cannot supply. The ethical question is not whether the past may be studied. It is who authorises the destruction, who bears the loss, who controls the resulting data, and who shares in the knowledge produced. A field built from irreplaceable ancestors carries those questions inside its method, not outside it.

How to Read an Ancient DNA Claim: Evidence Classes in Order

A careful reader begins by asking what material was sampled, because material limits every later inference. A headline may announce a migration, a royal relationship, or an extinct lineage, yet the underlying evidence may rest on a complete genome, a short mitochondrial segment, loose sediment, or a protein screen. Those categories are not interchangeable grades of the same result. Each answers a different kind of question, and confusion begins when a finding from a narrow evidence class is reported in the language of a broader one. The table below settles that first distinction by placing six common evidence lines in order of the material behind them, the claim each can support, and the limitation that must travel with the claim.

Evidence line	Material sampled	What it can establish	Main limitation
Whole genome from petrous bone	Dense inner ear bone	Broad ancestry, kinship, sex and many individual variants	Destructive sampling and uneven preservation across regions
Genome from tooth dentine	Dentine beneath tooth enamel	Genome wide ancestry and kinship with often moderate coverage	Lower yield than petrous bone and variable microbial load
Mitochondrial DNA only	Any preserved tissue yielding mitochondrial copies	Maternal lineage placement and contamination screening	One inherited line and no genome wide ancestry estimate
Sediment DNA with no skeleton	Cave or lake sediment layers	Presence of species and broad ecological or population turnover	No identified individual and risk of movement between layers
Pathogen genome from tooth pulp or calculus	Pulp chamber or mineralised dental plaque	Infection identity, pathogen evolution and some outbreak relationships	Presence in mouth or blood does not prove cause of death
Protein screening of unidentified fragments	Small bone or tooth fragments	Species or family level identification for selecting fuller analysis	Protein groups can be broad and give no genome sequence

Order inside the table reflects breadth of individual inference rather than importance. Sediment evidence can transform understanding of an ecosystem even though it names no individual, as Kjaer et al. demonstrated in Nature in 2022 with Greenland sediment material dated to about two million years ago. Protein screening can identify a fragment from a mixed assemblage and direct scarce sampling resources toward the right specimen. Pathogen genomes can rewrite the history of a disease without describing the ancestry of the person who carried the infection. A narrow result can be excellent science when its claim stays narrow.

The next questions concern authentication, coverage, and comparison. A strong report names the sampled element, fragment length, damage treatment, contamination estimate, coverage across the genome, and the reference data used for comparison. It also states whether another extract or laboratory supported the result. Van der Valk et al. in Nature in 2021 recovered mammoth genomes dated between roughly one million and 1.65 million years from Siberian teeth, a result whose force came from the combination of geological context, molecular damage, and comparison across specimens rather than from age alone. By contrast, a claim that withholds sample provenance or presents only a summary tree asks readers to transfer trust from stronger work. Evidence classes keep that transfer visible, and they give skepticism a precise target. Each added genome narrows uncertainty only when its provenance remains fully documented.

Why a Damaged Molecule Changed the Study of the Past

A damaged molecule changed historical inquiry because it introduced a witness that did not depend on writing, monuments, or deliberate preservation. Texts record what literate institutions chose to preserve, and artefacts record what materials and burial conditions allowed to survive. Genetic fragments add a different archive, one carried inside bodies and shed into caves, teeth, sediments, and burial soils. That archive is fragmentary, chemically altered, and difficult to authenticate, yet it preserves relationships that no chronicler thought to record. Parents and children, migrants and neighbours, infected hosts and pathogens, and populations that left no named history can enter the record through measurable biological connection rather than through later tradition.

The change was not simply the addition of more evidence. Genetic evidence reorganised the questions that archaeology and history could ask. A cemetery could be tested for biological kinship rather than assumed to mirror household structure. A stylistic change in pottery or burial practice could be compared against population continuity instead of being treated as proof of replacement or adoption on its own. An extinct relative could be identified from a bone too small to classify by shape. Mammoth genomes older than one million years and Greenland sediment material near two million years old extended biological evidence into depths where written analogy has no purchase. Each extension depended on restraint as much as ambition, because age and novelty increased the burden of authentication.

That burden explains why method became part of the historical result. A sequence without provenance, controls, damage assessment, and transparent data would add noise disguised as certainty. The field matured when laboratories learned to report absence, contamination, and uncertainty beside discovery, and when critics forced a distinction between a molecule that had been measured and an interpretation that had been attached to it. The strongest findings survived that pressure because independent lines of evidence converged. A migration claim gained force when genetics, chronology, material culture, and geography pointed in compatible directions. A kinship claim gained force when coverage and replication supported the precise relationship named. Evidence weakened when any one line was asked to do all of the work.

Cultural consequences followed from that discipline. Genetic ancestry can unsettle national myths, restore attention to neglected populations, and show movement and mixture where older histories imagined sealed groups. It can also be misused when statistical clusters are renamed as peoples, when a percentage becomes an identity, or when a sampled individual is made to speak for a continent. The mature lesson cuts in both directions. Biological connection is real evidence about the past, and it is not a complete account of belonging, language, law, memory, or worth. Holding both parts together prevents a powerful technique from shrinking human history to the molecules that happen to survive.

The evidence table in the preceding section offers the practical form of that lesson, sorting claims by material and limitation before narrative begins. Readers who keep that order in mind can welcome a genome, a sediment sequence, or a pathogen result at its proper strength without granting any of them unlimited authority. A short shelf of aids, including a reference tool for checking terminology and sources, can support that habit without replacing primary reports. The damaged molecule endures as a historical source because skepticism taught the field how to listen to it, and because its fragments, read within their limits, connect lives that documents never joined.

Frequently Asked Questions

Q: What is ancient DNA in simple terms?

Ancient DNA is genetic material recovered from organisms that died long before laboratory analysis, including humans, animals, plants, and microbes preserved in bone, teeth, hair, seeds, and sediment. After death, the molecule breaks into short pieces, sustains chemical damage, and mixes with material from soil organisms, so a recovered sample holds only a damaged trace of the original genome rather than an intact copy. Typical fragments measure only 30 to 70 base pairs, far shorter than the long strands obtained from living tissue. Researchers extract that trace under stringent clean conditions, convert fragments into sequence libraries, and compare patterns against reference genomes to infer ancestry, relationships, traits, and population movement across centuries and millennia with contamination controls applied throughout processing.

Q: How much bone is needed for an ancient DNA sample?

The amount depends on preservation, skeletal element, and laboratory protocol, and skilled teams often work with a surprisingly small quantity. Many extractions begin with about 50 to 200 milligrams of bone powder, roughly a pinch to a small spoonful, drilled from a dense interior region and weighed before chemical processing. Better preserved material from a cold or dry burial may yield usable sequences from the lower end of that range, while poorly preserved material may fail even when several hundred milligrams are consumed. Because sampling destroys the portion removed, curators and researchers weigh scientific value against preservation of the specimen, photograph and scan the element before drilling, select the least visible location, and retain powder or extract for replication where permission allows.

Q: Why is the petrous bone used so often in ancient DNA research?

The petrous portion of the temporal bone, a dense pyramid of bone deep inside the skull surrounding the inner ear, preserves genetic material far better than most skeletal elements. Its exceptional hardness and low remodeling during life limit water movement, microbial attack, and cracking after burial, so fragments inside survive longer and in higher concentration. Gamba et al. in 2014 demonstrated substantially higher endogenous yields from petrous samples than from teeth or other bones in the same individuals, and Pinhasi et al. in 2015 confirmed the advantage across a larger series. Sampling targets the densest inner region through the ear canal or a cranial opening where possible, reducing visible damage while obtaining the protected core that holds the richest molecular archive available.

Q: What is the oldest DNA ever recovered?

The oldest authenticated sequences come not from human remains but from exceptionally cold contexts that slowed molecular decay for immense spans of time. Van der Valk et al. in Nature in 2021 reported genomes from Siberian mammoth teeth dated between roughly one million and 1.65 million years old, recovered from permafrost that kept fragments frozen since burial. Kjaer et al. in Nature in 2022 reported sediment DNA from northern Greenland dated to about two million years old, preserved in frozen sediment and representing a whole ecosystem rather than a single organism. Both records depended on extremely short and damaged fragments, careful authentication, and geological dating of the surrounding deposits, and both lie far beyond the survival range expected in warm or wet environments.

Q: How do scientists determine the sex of ancient remains from DNA?

Sex estimation reads the balance of sequences mapping to the sex chromosomes rather than relying on skeletal shape, which works even for infants and fragmentary remains where anatomy gives no answer. A genome with two X chromosomes produces almost no Y specific sequences, while a genome with one X and one Y produces a characteristic proportion of Y reads alongside a lower X coverage relative to the autosomes. Researchers calculate ratios of X and Y coverage, compare them with thresholds established from individuals of documented sex, and report an assignment only when coverage is sufficient to separate the categories cleanly. Low coverage, contamination, and chromosomal variations can blur the signal, so responsible reports state the ratio, the coverage, and the confidence behind the assignment given.

Q: Can ancient DNA show whether two skeletons were related?

Kinship analysis compares genomes at many variable positions and measures how much more two individuals share than unrelated members of the same population would share by chance. Parents and children share one allele at nearly every position, siblings share long stretches in a pattern distinct from that of first cousins, and each step of distance reduces the shared fraction in a predictable way. Software estimates those sharing patterns from low coverage data by using genotype likelihoods rather than firm calls, which allows work on damaged material. Interpretation requires a suitable population background, because shared ancestry within a small or endogamous community raises baseline sharing for everyone, and contamination or very low coverage can create false closeness or hide a genuine relationship entirely.

Q: What proportion of an ancient DNA library is actually human?

In most extracts the human fraction is small, often well below one percent and sometimes only a tiny trace among millions of microbial sequences. Soil bacteria, fungi, and microbes that colonized the body after death contribute the overwhelming majority of molecules in a typical library, because burial environments teem with living and dead microorganisms while the host genome decayed into scarce short fragments. Exceptional petrous samples from cold or dry sites occasionally reach tens of percent endogenous content, and rare specimens exceed half, but such yields are the exception rather than the rule. Reporting the endogenous percentage matters because it determines sequencing cost, achievable coverage, and whether enrichment is worthwhile before deeper sequencing begins on a given specimen.

Q: Why do laboratories use negative controls in ancient DNA work?

Negative controls are blank samples carried through every laboratory step alongside genuine specimens, containing reagents and water but no bone, tooth, or tissue at all. If sequences appear in a blank, contamination entered during extraction, library preparation, or handling, and the batch cannot be trusted until the source is found. Because ancient material contains so few authentic molecules, even a trace of modern contamination from skin, breath, clothing, or a previous sample can outnumber the genuine signal and produce a plausible but false result. Laboratories therefore track blanks through sequencing rather than discarding them early, compare their contents with sample results, and halt reporting when a blank carries the same profile attributed to a specimen under study.

Q: What is mitochondrial DNA and why was it studied first?

Mitochondrial DNA is the small circular genome housed in mitochondria, the energy producing structures present in hundreds or thousands of copies inside most cells, inherited almost entirely through the mother. It was studied first because abundance solved the central problem of scarcity: each cell holds many mitochondrial copies but only two copies of most nuclear sequences, so degraded extracts were far more likely to retain a readable mitochondrial fragment. Early work therefore sequenced short mitochondrial regions to place remains within maternal lineages, compare populations, and identify species. Its limits are equally clear, since a single maternal line represents one ancestor among many thousands and records nothing about the rest of the genome, so nuclear data later became essential for complete ancestry and kinship analysis overall.

Q: How did ancient DNA change understanding of Neanderthal ancestry in living people?

Before genome scale work on archaic remains, many researchers expected little or no lasting contribution from Neanderthals to later populations, partly because mitochondrial lineages in living people showed no Neanderthal match. Green et al. in Science in 2010 compared a draft Neanderthal genome with genomes from living people across regions and estimated Neanderthal ancestry between 1 and 4 percent in genomes of people outside Africa, with no comparable signal inside Africa in that analysis. That pattern supported interbreeding after the expansion out of Africa rather than deep shared structure alone. Later higher coverage genomes refined the boundaries, identified surviving segments, and linked some segments to immunity, skin, and metabolism, changing the question from whether mixing happened to how its legacy was filtered by selection over time.

Q: Who were the Denisovans and how were they identified from DNA?

The Denisovans are an archaic population recognized from molecules before any diagnostic skeleton was found, which made the discovery a landmark for the field. A small finger bone from Denisova Cave in Siberia yielded mitochondrial sequences reported by Krause et al. in 2010 that matched neither Neanderthals nor modern humans, and the nuclear genome reported by Reich et al. in 2010 confirmed a distinct lineage closely related to Neanderthals. A molar from the same cave supported the finding, while later work identified a mandible from the Tibetan Plateau and traced Denisovan ancestry in living populations, strongest in parts of Oceania. Because so few fossils are assigned with confidence, most understanding about the group still comes from genomes rather than anatomy alone.

Q: Can ancient DNA identify diseases that killed people in the past?

Pathogen genomes recovered from teeth and bone can identify infections present at death, because blood borne microbes leave molecules in dental pulp and vascular tissue that survive alongside host material. Bos et al. in Nature in 2011 reconstructed the genome of Yersinia pestis from Black Death burials in London and confirmed that the bacterium responsible for plague caused that epidemic, settling a long debate sustained by circumstantial accounts. Comparable work has identified tuberculosis, hepatitis B, smallpox, and other infections in earlier populations, sometimes centuries before written descriptions. Detection shows presence rather than certainty about cause of death for an individual, since chronic infection, coinfection, and postmortem contamination must all be considered before a diagnosis is claimed for a particular burial context.

Q: Does ancient DNA work on cremated remains?

Cremation is among the harshest treatments a body can undergo, because sustained high heat fractures molecules, drives chemical change, and often leaves bone white, warped, and brittle. Most cremated remains yield no usable endogenous sequences at all, especially after prolonged burning at pyre temperatures, and laboratories treat any result from such material with heightened caution. Occasional successes come from incompletely burned elements, protected interior fragments, or teeth that escaped the hottest zone, where a small pocket of material survived below the threshold of complete destruction. Researchers therefore test cremated samples only with modest expectations, report failures openly alongside successes, and avoid drawing population conclusions from the rare positive result that an unusual burial happened to preserve by chance.

Q: Why does heat destroy ancient DNA faster than cold?

Heat accelerates the chemical reactions that break genetic molecules, while cold slows those same reactions and preserves fragile bonds for far longer. Warmth speeds hydrolysis that snaps the backbone, encourages oxidation, and supports active microbes and enzymes that digest exposed material, and cycles of wetting and drying in warm soil multiply the damage. Frozen, dry, and stable cold environments suppress microbial growth, reduce water available for hydrolysis, and hold reaction rates low across thousands of years, which explains why permafrost and high latitude or high altitude sites dominate the oldest records. Temperature alone does not decide survival, since acidity, water movement, and mineral chemistry also matter, but sustained warmth is consistently the strongest predictor of poor preservation in comparative surveys.

Q: What is DNA capture enrichment and when is it used instead of shotgun sequencing?

Shotgun sequencing reads everything in an extract indiscriminately, host and microbe alike, which is thorough but wasteful when the host fraction sits below one percent. Capture enrichment takes the opposite approach, using manufactured probes that bind selected targets and pull them from the library before sequencing, so effort concentrates on informative positions such as a defined panel of variable sites or the mitochondrial genome. Laboratories choose capture when endogenous content is low, when many samples must be screened at shared positions for population comparison, or when a pathogen genome is the target. Capture sacrifices discovery outside the probe set and can bias allele representation, so shotgun sequencing remains preferable when preservation is strong and unbiased whole genome coverage is the goal sought.

Q: Can hair preserve ancient DNA without a root?

Hair shafts without roots contain very little nuclear material, because the shaft consists mainly of keratin and the cells that formed it lost their nuclei during growth. For many years that structure led researchers to treat rootless hair as a source for mitochondrial sequences only, since mitochondrial copies persist in the shaft in higher numbers. Specialized extraction methods that dissolve keratin thoroughly and recover minute fragments have since obtained nuclear genomes from historic hair, including material preserved in museums without any root attached. Success remains inconsistent and depends on chemical treatments applied during life, washing, dyes, and storage conditions, so a failed extraction from hair does not prove that no recoverable material was present in the original specimen at all.

Q: How accurate is ancestry reconstruction from a single ancient genome?

A single genome can establish broad ancestry with considerable confidence, placing its owner within major population groupings and detecting substantial archaic or distant contributions when coverage is adequate. Finer claims are far weaker, because one individual cannot represent a whole population, and low coverage, damage patterns, and contamination narrow the positions that can be read reliably. Reconstructions also depend on comparison panels drawn largely from living people, whose structure may map imperfectly onto past populations that left no direct descendants. Responsible interpretation therefore reports broad placement firmly, treats precise percentages and trait predictions cautiously, checks results against archaeological context, and waits for additional genomes from the same place and period before describing a population rather than a person alone.

Q: What happens to ancient DNA results when a reference genome changes?

Ancient results are produced by aligning short damaged fragments against a reference genome, so the reference acts as a coordinate system and a filter for everything reported. When a revised reference appears, with corrected assemblies, added sequences from diverse populations, or a pangenome replacing a single linear standard, realignment can shift variant calls, recover fragments that previously failed to map, and alter ancestry estimates at the margins. Well supported findings based on abundant data rarely reverse, because their signal rests on many independent fragments, but borderline calls and regions absent from an older reference can change materially. Careful groups archive the reference version used, publish alignment files, and reprocess key samples when a major reference change could affect a published conclusion in practice.

Q: Why do some famous ancient DNA claims get withdrawn or disputed?

Famous claims attract dispute because early results in the field sometimes rested on short sequences obtained before clean laboratories, authentication criteria, and contamination controls reached mature standards. A sequence that matched a dramatic expectation could instead have come from a handler, a laboratory reagent, or a microbe, and later teams with better methods sometimes failed to replicate the finding in the same or comparable material. Journals and authors have withdrawn or corrected reports when contamination was demonstrated, when dating of the specimen changed, or when reanalysis showed that the signal belonged to another organism. Such corrections strengthened rather than weakened the discipline, since each failure produced stricter blanks, independent replication in a second laboratory, and damage pattern tests that genuine material must pass consistently.

Q: How is ancient DNA different from forensic DNA testing?

Forensic testing works with fresh material collected for identification, where long intact molecules allow standardized profiling at a defined set of markers for identification and court presentation. Ancient work begins with scarce fragments commonly 30 to 70 base pairs, heavy damage, and overwhelming microbial content, so methods emphasize authentication, contamination exclusion, and population level inference rather than a single definitive profile. Chain of custody and legal admissibility govern forensic practice, while destructive sampling permissions, archaeological context, and descendant community interests govern ancient research. The two fields share laboratory chemistry and respect for contamination, yet a result that satisfies a research question about ancestry across millennia would rarely meet the evidentiary standard required to identify a person in a legal proceeding.