HubG25: Admixture JS

Download datasets

Open files

gradient palette

nearfar

gradient palette

nearfar

add dist colno

off0.5x1x1.5x2x

weight conditions i

Mode

Custom population name

Type a population name to fetch its coordinates from your SOURCE/TARGET data, or paste a G25 line. Set a weight (%), then add or subtract — one population at a time. Additive weights must sum to 100%; subtractive weights must total under 100%. Weights and signs stay editable in the list below.

Population

Weight %

Additive total: 0% (need 100%)

Output

Ghost Generator solves T = Σ(addᵢ·Aᵢ) − Σ(subⱼ·Sⱼ) for the residual population, rescaling by 1/(1 − Σsubⱼ) so the mixture coefficients sum to 1. Weighted Average Mode builds sample-size–weighted means labelled Name_(n=N).

01 · Targets

One target per line: label, then comma-separated coordinates. Click Load targets to build the list.

02 · Sources

One population per line. At least two. Click Load sources to refresh the checklist.

03 · Bootstrap replicates

Dimensions resampled with replacement; each replicate is refit. Applies to every target.

Each target is fitted against the same checked sources, and every target gets its own result block below. Fit is the NNLS solution on all dimensions. SD is the spread across bootstrap resamples. Z = Fit ÷ SD — high Z means load-bearing, low Z means noise.

Stability results

Paste one target per line and your sources, load both lists, then run all or run a single target. Z ≥ 3 is stable; 1–3 provisional; < 1 noise.

Disclaimer: this tab's baseline Fit % is the exact NNLS least-squares solution on all G25 dimensions. Single and Multi admixture instead use a fast, stochastic Monte Carlo optimizer (their cycles setting) that approximates that same solution rather than solving it directly, so their weights and distance can come out slightly different — usually by a fraction of a percent. Raising cycles in Single/Multi narrows the gap at the cost of speed.