Methodology deep-dives, population spotlights, and engineering notes from the Ancestrify team, qpAdm modelling, ancient DNA panels, and how genetic ancestry analysis really works.
24 stories
Iberia has the longest continuous ancient-DNA transect in Europe: foragers, farmers, a Beaker-era Y-chromosome revolution, Phoenicians and Romans, the Islamic centuries and their aftermath. What each layer left, and how to read an Iberian genome.
Ireland's ancient-DNA transect is one of Europe's cleanest: a Mesolithic baseline, a Neolithic of tomb-builders with a dynastic elite, a near-total Beaker-era turnover, and striking continuity since. The story, and how to read an Irish genome.
Germany is ancient DNA's best-sampled territory: the LBK farmers, the Corded Ware steppe arrival, Bell Beakers, and the Celtic–Germanic–Slavic interfaces all run through it. The layered story, and how to read a German genome.
Poland's ancient-DNA story has a twist most national histories lack: a documented population turnover in the first millennium CE. Goths and Wielbark, the early Slavic horizon, what modern Polish genomes show, and how to read your own.
Iran holds one of ancient DNA's founding populations. What the Zagros genomes changed, how the plateau's ancestry formed and persisted through empires, what modern Iranian samples show, and how to read your own results.
Ancient DNA gives Armenians one of West Asia's clearest stories: a Bronze Age blend of local Caucasus and Anatolian streams, a measured steppe pulse, and genetic isolation since roughly the end of the Bronze Age. The evidence, honestly read.
Georgia holds two of the most important genomes in West Eurasian prehistory. What Satsurblia and Kotias revealed, the Kura-Araxes and later layers, why Georgian ancestry is among Eurasia's most continuous, and how to read your own results.
Kurdish ancestry through the ancient-DNA lens: the Zagros Neolithic foundation, the layers that followed, what modern samples show about structure and neighbours, and how to read G25 and qpAdm results for a Kurdish genome.
What ancient DNA and modern genetic studies say about Assyrian ancestry: deep Mesopotamian roots, two millennia of documented endogamy, the missing ancient transect, and how to read G25 and qpAdm results for an Assyrian genome.
f2, f3, f4 and D-statistics are the shared-drift arithmetic beneath qpAdm, qpWave and admixture graphs. What each statistic measures, how a four-population test works, and how to read Z-scores like the papers do.
From genotype files to a tested model: f2 extraction, qpadm() and its output tables, the arguments that silently change results (allsnps, fudge_twice, constrained), and the protocol discipline the code will not enforce for you.
Negative weights, SE 9.99, every model rejected, every model passing, infeasible popdrop rows, allsnps confusion — the standard failure gallery of qpAdm runs and what each symptom actually indicates.
qpAdm's sibling asks a prior question — how many independent streams of ancestry does a set of populations need? The rank test, cladality checks, and why every good qpAdm search starts with a qpWave answer.
The canonical nine outgroups from Lazaridis 2016, the o9a and o9aamcn extensions, the arithmetic floor and the ~30-population ceiling, and the published example where one added outgroup cut standard errors threefold.
The rotating protocol tests every candidate as source and outgroup in turn — elegant, recommended by the method's auditors, and carrying a measured 72–100% false-discovery rate when run without temporal discipline. What rotation is actually for.
Whether your sources predate your target is not a style preference — measured false-discovery rates run 16–31% for temporally stratified protocols and 72–100% for proximal rotating screens. What each model type is for.
The number that governs a qpAdm model is never your file's marker count — it is the intersection with the ancient panel, per statistic. The published coverage numbers, the allsnps table, and what a consumer chip can honestly support.
ADMIXTURE describes structure with K inferred components; qpAdm tests an explicit historical model and can reject it. What each estimates, why papers use both, and which answers which of your questions.
qpAdm first appeared in the supplement of the 2015 steppe-migration paper and became the method behind a decade of ancestry headlines. Where it came from, what it settled, and how a decade of stress-testing sharpened its limits.
The fit distance measures how far a mixture still sits from your coordinate — not whether the model is right. Calibrated bands for reading one, why lower stops being better, and the checks that actually validate a model.
Every Global25 row exists in two forms, and mixing them is the hobby's most common silent error. What scaling does mathematically, which form each tool expects, and the tell-tale signs of a mixed comparison.
Adding sources always lowers a G25 fit distance and routinely worsens the model — a worked demonstration where the fit improved from 0.0110 to 0.0068 while the story collapsed. The rules that keep a panel honest.
Global25 is a projection, not a test — so 'accuracy' has layers: the coordinate's own fidelity, the references it is compared against, and the models built on top. An honest audit of all three.
Ger Huijbregts's nMonte turned Global25 rows into ancestry percentages and founded a whole tool tradition. How the Monte-Carlo search works, what nMonte3 changed, what the fit distance is, and the discipline the script never enforces.