Reading Bones Like Books

Core Concepts

1. What Ancient DNA Actually Is

Ancient DNA (aDNA) is genetic material recovered from archaeological remains — bone, teeth, occasionally preserved soft tissue. Unlike modern DNA, it has spent years to millennia exposed to the chemical forces of decomposition. What survives is fragmentary, chemically altered, and often vastly outnumbered by environmental DNA from bacteria, fungi, and soil organisms that colonized the remains.

This is not a minor technical inconvenience. A single ancient sediment sample contains DNA from a broad range of species mixed together, which heavily restricts what can be recovered and how. The human signal — the thing researchers are looking for — is typically a tiny fraction of total sequenced reads.

The field has grown enormously despite these constraints. The Reich Lab at Harvard alone has analyzed over 15,000 ancient genomes, combining data from 10,016 ancient individuals with 5,820 previously published sequences and 6,438 modern genomes, covering 18,000 years of human history. What David Reich himself described as an "industrialization" of ancient human population genetics has produced a database of unprecedented scale — though one concentrated heavily on Europe and the Middle East.

2. Getting DNA Out: Extraction Protocols

Recovering ancient DNA requires overcoming two opposing pressures: maximizing recovery of whatever endogenous DNA exists, while minimizing introduction of new contamination and further degradation of the sample.

Standard protocols use bone powder treated with EDTA and proteinase K buffers, followed by purification through silica binding in guanidinium thiocyanate solutions — all performed at room temperature to avoid heat-induced damage. Teeth and petrous bone (the dense inner ear bone) tend to yield the best preservation, as their compact structure shields DNA from soil microbes. Rare soft tissues require adapted approaches.

For large collections, high-throughput extraction methods using automated robotic platforms allow researchers to screen extensive museum archives and archaeological deposits to identify promising samples before committing to full sequencing. This addresses a historical bottleneck: you can't know which samples will yield usable DNA until you try.

3. The Contamination Problem

Modern DNA contamination is a primary challenge in archaeogenomics, arising from researchers themselves, from laboratory aerosols, from postmortem soil microbes, and from prior handling of museum specimens. The math is stark: even a single molecule of modern human DNA can compromise analysis, because it competes directly with the small amounts of endogenous aDNA and is indistinguishable without further evidence.

This is why dedicated ancient DNA laboratories maintain physical separation from modern genetics work, enforce full-body protective equipment, bleach-decontaminate surfaces and tools, and apply UV irradiation to equipment. It also shapes the logic of authentication: you cannot trust a result unless you can demonstrate that what you sequenced is genuinely old.

4. Authentication Through Damage Patterns

The solution lies in the chemistry of decomposition itself. Ancient DNA carries characteristic post-mortem damage patterns that modern contaminants lack. Two mechanisms dominate:

Cytosine deamination: Cytosines convert to uracils over time, which read as thymine during sequencing. This produces characteristic C-to-T substitutions (and G-to-A on the complementary strand). These substitution rates reach up to 40% at fragment ends, dropping off exponentially along the fragment — a pattern that is predictable, measurable, and entirely absent in freshly extracted modern DNA.

Depurination-driven fragmentation: Purine bases detach from the sugar-phosphate backbone, causing the strand to break. The result is very short fragments, typically under 100 base pairs. When you sequence ancient remains and find overwhelmingly short fragments bearing terminal C-to-T damage, you have strong evidence you are looking at genuine ancient DNA.

These damage kinetics follow predictable mathematical models of decay. Tools like mapDamage quantify these patterns computationally, turning a chemical signature into a statistical authentication check.

The very thing that makes ancient DNA difficult to work with — its damage — is also what proves it is real.

5. Community Standards: SPAAM

Reliable archaeogenomics requires more than good lab technique. It requires consistent metadata, reproducible protocols, and shared quality standards across labs. The SPAAM community (Standards, Precautions, and Advances in Ancient Metagenomics) maintains exactly these: MInAS standards (Minimal Information for Ancient Samples), a curated AncientMetagenomeDir database of published ancient metagenomic samples with standardized metadata, and training resources including a 2024 textbook on ancient metagenomics. Without such infrastructure, comparing results across labs or integrating datasets becomes unreliable.

6. Computing Ancestry: qpAdm and PCA

Once you have authenticated ancient DNA sequences, the next challenge is inferring what they reveal about population history. Two tools dominate this work — and both carry systematic biases worth understanding.

qpAdm is a statistical framework for modeling how much of a target population's ancestry derives from each of several proposed source populations. It has been essential for reconstructing admixture events across prehistory. But systematic performance problems emerge under realistic ancient DNA conditions: when reference panel populations are suboptimal or incomplete — a persistent reality for under-sampled global regions — admixture weight estimates skew toward whichever reference populations are most similar to the true sources. The method also tends to favor geographically distant sources over nearby populations on simulated landscapes. And critically, when researchers test hundreds to thousands of competing models (a common practice), false discovery rates can exceed 50% in substantial portions of parameter space.

PCA (Principal Component Analysis) is the dominant visualization tool — the scatter plots that position ancient samples relative to modern populations. It offers an intuitive picture of genetic similarity, but contains systematic biases that have led to widespread misinterpretation. PCA results are artifacts of data composition and reference sample selection: the same samples plotted against different modern reference panels will land in different positions. Wave-like patterns that look like migration routes can arise from simple spatial genetic decay rather than historical movement. Visual clustering can obscure underlying demographic complexity.

Neither tool is wrong to use; both require demographic modeling as a complement rather than treating visual results as self-evident.

7. Archaic Admixture: Neanderthals and Denisovans

One of ancient DNA's most consequential discoveries is that modern humans interbred with archaic hominin populations — and some of those variants persist in living people because they were useful.

The introgression was not a single event. Evidence shows at least three introgression events from distinct Denisovan populations, each presenting different levels of relatedness to the sequenced Altai Denisovan individual. Neanderthal gene flow occurred recurrently over roughly 200,000 years, across multiple geographic regions and time periods.

Not all introgressed variants survived. The ones that did often did so because selection favored them. 126 high-frequency archaic haplotypes show signatures of adaptive introgression, significantly enriched for immune-related and skin pigmentation genes. The OAS gene cluster (OAS1/2/3), which encodes components of the antiviral response pathway, provides a well-documented case: Neanderthal haplotypes at this locus reach frequencies above 40% in Eurasian populations while remaining nearly absent in sub-Saharan Africa. Modern humans who interbred with Neanderthals — who had lived in Eurasia for approximately 200,000 years before the modern human dispersal — appear to have acquired pathogen defenses already calibrated to Eurasian environments.

The Takarkori individuals from the Green Sahara (7,000 years old) offer a useful comparison point: they carried approximately 0.15% Neanderthal DNA — about 10 times less than non-African populations (0.6–0.9%), but significantly more than contemporary sub-Saharan African genomes. The gradient itself is informative about where and when gene flow occurred.

8. Sex-Biased Admixture and What Genetics Can and Cannot Say

Ancient DNA can often distinguish male and female ancestry by comparing autosomal (non-sex chromosome) patterns against Y-chromosome and mitochondrial signals. This enables detection of sex-biased admixture — where one sex contributed disproportionately to a mixed population.

The Papuan admixture into Remote Oceania is an instructive example. Ancient DNA identified individuals with Papuan male lineages alongside Austronesian female ancestry, consistent with sustained inter-group contact along trade networks rather than wholesale population replacement. The admixture began around 2,500 years before present and proceeded progressively over centuries — DNA segment lengths from Papuan ancestry are longer and less recombined in earlier individuals, breaking down across generations as expected under an extended admixture model. The source populations traced to the Bismarck Archipelago and Solomon Islands.

This same study demonstrates the critical interpretive boundary: in Remote Oceania, Austronesian languages persisted even as genetic ancestry was substantially replaced by Papuan ancestry. Language continuity and genetic continuity are independent. Genetics can tell you about biological ancestry and demographic movement. It cannot tell you what people called themselves, what language they spoke, or what cultural practices they carried. These require independent evidence.

9. The Three-Ancestry Model of Europe (and Its Limits)

One of the most widely cited results of European archaeogenomics is that modern Europeans descend from three major ancestral populations: Mesolithic hunter-gatherers, Anatolian Neolithic farmers, and Pontic-Caspian steppe pastoralists. This tripartite model has been productively used to explain shifts in ancestry proportions through time.

Ancient DNA shows that the admixture between these groups was not instantaneous. Early Neolithic farmer populations showed minimal admixture with local hunter-gatherers, but subsequent populations increasingly incorporated hunter-gatherer ancestry — indicating prolonged coexistence and gradual biological integration rather than rapid replacement. Regional timing varied substantially.

The "ancestry trinity" is analytically useful but should be understood as a simplification. The three components are themselves composite, and the model's apparent clarity partly reflects the Eurasian focus of available reference panels.

Common Misconceptions

"Ancient DNA tells us where people came from." It tells you about biological ancestry — which populations contributed genomic material to the individuals sampled. Migration, trade, contact, and cultural transmission leave different signatures, and these don't always co-occur. A person can carry the genetics of a distant population without having moved, if their ancestors did.

"More Neanderthal DNA means more connection to an ancient past." Neanderthal ancestry in non-African populations (roughly 1–4%) is not a sign of primitiveness or distinctiveness — it reflects a historical contact event that happened to yield selectively advantageous variants that stuck around. Sub-Saharan populations that have less Neanderthal ancestry are not "more modern"; they simply descend from populations that had less contact with Neanderthals geographically.

"PCA plots show where ancient people lived." PCA plots position samples in a mathematical space defined by genetic similarity to reference populations. Geographic clustering in these plots reflects shared ancestry patterns, not geographic location at death. The same sample can appear to "move" depending on which reference populations are included.

"qpAdm results are the accepted answer." qpAdm outputs are model-dependent estimates, not measurements. They depend heavily on which reference populations are chosen, how many models are tested, and how well-sampled the relevant ancestral populations are. High false discovery rates when testing large model sets are a documented property of the method, not an edge case.

"Ancient DNA and cultural remains tell the same story." They frequently don't — and the divergences are informative. Language, material culture, genetic ancestry, and social identity are separate streams that can and do decouple. The Remote Oceania case, where language persisted across near-complete genetic replacement, is one documented example.

Boundary Conditions

Where ancient DNA preservation fails. aDNA degrades rapidly in warm, humid conditions. Tropical regions — which include enormous swaths of human prehistory — are dramatically under-represented in current databases, not because they're less important but because preservation conditions are far worse. The geographic skew of the current dataset is partly a physics problem, not only a funding or priority problem.

When reference panels are too thin. Both qpAdm and PCA require reference populations that adequately represent ancestral sources. Where those sources are unknown, absent from sequenced databases, or poorly represented, inferences become unreliable in ways that may not be immediately visible in the output. This is a particular problem for Africa, South and Southeast Asia, and the Americas — all of which are dramatically under-sampled relative to their historical significance.

The resolution limits of aDNA. aDNA is sequenced at varying coverage depths. Low-coverage data (a common product of degraded samples) can support some inferences about ancestry but not fine-grained analysis of individual variants. Single-nucleotide polymorphism (SNP) capture methods recover specific positions rather than whole genomes, which is efficient but means you're only looking at pre-selected sites.

Ethics and community consent. Ancient human remains frequently belong to living descendant communities. The archaeogenomics field has faced substantive criticism over sampling practices that proceeded without adequate consent or collaboration — particularly for Indigenous communities. SPAAM standards and community-led frameworks are responses to this, but the field is still developing appropriate norms. Findings about ancestry have real-world implications for community identity and political rights, which adds stakes well beyond academic interpretation.