Strataa research instrument for population genetics
Abstract
Strata places a genome, or a set of Global25-style coordinates, among a curated library of ancient and present-day reference populations. It projects samples into a shared principal-component space, ranks nearby populations, fits exploratory mixture models with bootstrap intervals, and runs formal qpAdm tests on f-statistics through ADMIXTOOLS 2. Every result states its method, its uncertainty and the reference data it rests on.
1.What it does
Loading the reference space…
- Project. Place a sample in the reference PCA using only the SNPs it carries.
- Compare. Rank reference populations by distance in the same space.
- Model. Fit and rotate candidate sources, then test the surviving models formally with qpAdm.
- Ask. Pose questions to a research assistant that runs these analyses and cites what it ran.
2.Exploratory fits and formal tests
Strata keeps two families of results apart, because they answer different questions. The distinction is marked on every result.
| exploratory | formal | |
|---|---|---|
| Method | Non-negative least squares on PCA or G25 coordinates | qpAdm on f4-statistics (ADMIXTOOLS 2) |
| Question | Which mixture of sources lies closest to the target in this space? | Is the target consistent with descending from these sources, given outgroups? |
| Uncertainty | Bootstrap over source individuals; sensitivity to dropped sources and dimensions | Block-jackknife standard errors; tail probability of the rank test |
| Reading | A description of geometry. A good fit is not evidence of admixture. | A model can be rejected. A model that is not rejected may still be wrong. |
3.Worked examples
Each example is computed from public reference data with the same code you run here, and carries its inputs and provenance.
Loading examples…
4.Where to begin
Browse the reference library to see which populations and periods are covered. Then upload a raw file from 23andMe, AncestryDNA, FamilyTreeDNA or MyHeritage, a single-sample VCF, or paste G25 coordinates, and compose an analysis.
The methods describe each statistic and its assumptions.