Rotation

When you run a standard admixture model, you pick one specific set of sources and test that combination. Rotation automates the search: you give it a pool of candidate populations and it tests every combination up to K sources, ranking the results by how well each one fits your DNA.

A rotation sweep can test hundreds of combinations in a single run and return only the ones that pass the statistical threshold, so you don't have to enumerate the possibilities manually.

K and the source pool

K is the number of sources per model (2, 3, or 4). Every combination of K populations drawn from your pool is tested as a separate qpAdm model. A pool of 8 populations with K=3 produces 56 combinations.

You can also split the pool into competing and fixed sources. Competing sources are drawn from the pool in different combinations. Fixed sources appear in every model unchanged. Fixed sources are useful for populations you already know must be present, leaving the pool to fill in the unknown components.

Fixed references

Fixed references are the outgroup populations used to calibrate the comparison, the same concept as the right populations in a single qpAdm run. The defaults work for almost every modern sample and are a reasonable starting point.

Reading the results

Each row in the results table is one tested combination. The columns show which source populations were used, the weight each was assigned as a percentage, and the p-value. A p-value above 0.05 means the model passes. The table is sorted with passing models first, then by p-value descending.

• Focus on models where all weights are positive and the percentages sum close to 100%. Models where one source gets nearly 0% are usually redundant.
• A passing p-value doesn't guarantee the model is historically correct. Use your knowledge of the populations to filter out implausible combinations.
• If no models pass, try reducing K, adding more populations to the pool, or checking your fixed references.

Model cap

Rotation is capped at 1,600 models per run. If your K and pool size would generate more combinations than that, the run will not start. The model count is shown in the configure panel as you adjust K and the pool. Reduce K or trim the pool to bring the count down.

All SNPs

By default rotation uses only genome positions covered across every population in the run. All SNPs mode uses every covered position and fills missing data with per-population averages, giving more statistical power at the cost of potential bias.

All SNPs applies to every model in the sweep. For well-covered modern kits, leave it off. Enable it only if the default sweep rejects all models and your target has low coverage.

Engine choice

Rotation offers two engines. ADMIXTOOLS2 is the default: it's faster and uses less memory, making it suitable for large sweeps.

Legacy AdmixTools uses the original DReichLab C binaries and is slower, but some researchers prefer it for direct comparability with published results that used the same software. For most users, ADMIXTOOLS2 gives equivalent results.

Credit cost

Rotation sweeps are included in your plan. Run from your weekly credit allowance, each sweep is a flat 10 credits (either engine), so a Basic plan covers about 50 rotations a week and Pro about 70. Once your weekly allowance is used up, rotations run on purchased credits at their list price: 4 credits ($2.00) on ADMIXTOOLS2 or 3 credits ($1.50) on legacy AdmixTools. The price is flat no matter how many models the sweep runs, up to the 1,600-model cap.