Genetic geography in colour
These maps revive the RGB “synthetic map” tradition. Each of the first three principal components is first converted into a separate intensity surface: PC1 controls red, PC2 green, and PC3 blue. The three surfaces are then superimposed pixel by pixel as the colour channels of one composite RGB map. Places with similar hues occupy similar positions across all three axes, while changes in hue let the eye follow multivariate clines, boundaries, and local deviations that would otherwise require comparing three separate maps. The composite is descriptive and model-dependent; it is not an ancestry calculator. Spain, the Europe-basis Great Britain view, and European Russia are country details of the Europe analysis. Britain also has a separate country-only PCA for greater within-country contrast. France and the Europe-inclusive Italy view use separate fine-geography bases.
Start with Europe ↓
Individual PCA distribution
The plot shows 2,363 individual reference scores on the same reliability-weighted Europe-only PCA coordinates used by the RGB map. Countries are shown as country-scale distributions, while Italy is divided only into Northern, Central and Southern Italy.
Hover over a hull, or select a population to isolate it.
Methods and boundaries
The Europe, France, and Europe-inclusive Italy bases fit reliability-weighted population means. The country-only Britain view fits every retained person simultaneously from its raw-genotype country panel while controlling each locality's total influence; nobody is held out.
PC1 becomes red, PC2 green, and PC3 blue. Europe, Spain, and European Russia use equal-population empirical percentiles, so an extreme population cannot consume most of a channel. Great Britain, France, and Italy use explicitly labelled local contrast scales to reveal within-country variation. Britain-only uses individual-score 2.5th–97.5th percentiles. The three surfaces are then superimposed pixel by pixel. Colours preserve ordering within a PCA basis but are not linear PCA distances. Separate PCAs can rotate or exchange axes, so their RGB hues are not directly comparable.
A mapped location is a population or locality mean placed at a disclosed geographic point; it may be an exact collection locality or a clearly labelled representative centre. These observed means are interpolated over land. A country view requires at least three mapped locations and uses only those locations.
Colours summarize PCA position and are not ancestry percentages. PC1 and PC2 are stable in the 30-replicate population-stratified bootstrap; Europe-only PC3 is exploratory and less stable. The Sámi PC2 shift persists after reliability weighting, but remains an axis-specific rather than whole-genome distance statement. The separate Italy fine-geography analysis displays the recovered Naples records alongside the audited Italian localities; Sardinia remains a disclosed population-level sensitivity anchor. Smooth transitions can reflect migration and isolation by distance, while sharp local contrasts may reflect drift or barriers—but the map alone cannot identify a historical cause.
Transparent calculation
Counts change when you switch analyses because each map has its own fitting panel and geographic eligibility rules. These categories are kept separate below.
Exact sequence
All eligible pre-cap references were intersected with the KING unrelated set, and all 348 curated Italian samples were added. Model-specific genotype QC excluded two all-homozygous encodings (HG00124.SG and NA20526.SG), leaving 2492 people in 38 European labels. All 107 1KG TSI are retained as a distinct fitting population.
Every person was retained. Individual weight was 1/max(population N, 30), so a population's total influence is min(1, N/30): cohorts above 30 cannot dominate, while tiny groups are not amplified to full-population leverage. Axes were fitted to one mean standardized-genotype vector per population using 8,180 shared, LD-pruned autosomal SNPs.
Italian, French, Western Balkan and British samples contribute through their controlled source-population groups; locality labels do not become extra PCA-fitting units. Fine geography is used for mapping wherever available. The 99 CEU references remain in the PCA but are excluded geographically, leaving 2393 people at 123 mapped locations.
Population centroids set the robust colour limits, preventing the finer Italian, French, Western Balkan and British locality grids from determining the scale. Locality means are then interpolated by inverse-distance weighting (power 2.5) on a 40 km land grid within 600 km of observed data, with spatial influence multiplied by min(n, 5)/5.
The 30-replicate population-stratified bootstrap gives median aligned correlations of 0.994 for PC1, 0.958 for PC2 and 0.652 for PC3. PC3 remains the least stable axis and should be treated as exploratory.
Population-level audit
PCA fitting totals include CEU; mapped totals do not.
Download CSV ↓| Population label | Scope | PCA fitting | Mapped | Mapping status |
|---|---|---|---|---|
| Albanian | Europe | 7 | 7 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Balearic | Europe | 25 | 25 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Balkan | Europe | 94 | 94 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Basques | Europe | 213 | 213 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Belarusian | Europe | 8 | 8 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Bulgarian | Europe | 51 | 51 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| CEU | Europe | 99 | 0 | Excluded: Utah CEU sample has no defensible single European locality |
| Corsica | Europe | 16 | 16 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Estonians | Europe | 34 | 34 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Finns | Europe | 116 | 116 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| French | Europe | 284 | 284 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| GBR | Europe | 91 | 91 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Greek | Europe | 111 | 111 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| IBS | Europe | 202 | 202 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Italy_C | Europe | 73 | 73 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Italy_N | Europe | 91 | 91 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Italy_S | Europe | 89 | 89 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Karelian | Europe | 17 | 17 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Lithuanian | Europe | 352 | 352 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Moldavian | Europe | 18 | 18 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Mordovian | Europe | 37 | 37 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Nordic | Europe | 3 | 3 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Orcadian | Europe | 15 | 15 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Polish | Europe | 22 | 22 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Portugal | Europe | 10 | 10 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Provence | Europe | 5 | 5 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Russian | Europe | 74 | 74 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Russian_Archangelsk_Krasnoborsky | Europe | 6 | 6 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Russian_Archangelsk_Leshukonsky | Europe | 5 | 5 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Russian_Archangelsk_Pinezhsky | Europe | 5 | 5 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Russians_Central | Europe | 3 | 3 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Saami | Europe | 23 | 23 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Sardinian | Europe | 28 | 28 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Swedish | Europe | 19 | 19 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| TSI | Europe | 107 | 107 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Udmurt | Europe | 10 | 10 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Ukrainian | Europe | 117 | 117 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
| Veps | Europe | 12 | 12 | Mapped with fine locality detail where available; otherwise at a disclosed regional centre |
Population names are source-dataset labels. Italy, France, the Western Balkans and Britain use fine localities where available; remaining samples use disclosed representative centres.
Open materials
All eligible unrelated references plus all 349 directly available raw Italian samples; France, Italy, the Western Balkans and Britain contribute through controlled source-population groups. Fine localities are mapping units only. Stored PC signs; equal-population empirical-percentile colour scaling for RGB and individual PC layers; sample-size-reliability-weighted IDW on a 40 km grid within 600 km of data (full locality weight at n = 5).
