SEPIA

Systematic Evaluation of Pigmentation In Archaic hominins

SEPIA — Systematic Evaluation of Pigmentation In Archaic hominins — asks what the genetics of pigmentation (skin, hair, and eye colour) can, and cannot, tell us about Neanderthals and Denisovans, and how they compare to present-day human variation. The recurring public question is “what did archaic humans look like?” SEPIA asks how much their ancient DNA can actually answer — and stays explicit about where the data runs out.

The headline result

Where the archaic hominins fall relative to modern humans depends entirely on which pigmentation variants you look at. We build a PCA on modern human genomes and project the archaics onto it in three ways — the whole genome, whole pigmentation-gene regions, and a hand-picked panel of pigmentation-associated SNPs. The whole genome and the pigmentation genes give the same answer: the archaics sit as a divergent outgroup, exactly as their deep ancestry predicts. Only the ascertained SNP panel places them “with Africans” — and that is a footprint of selection and ascertainment, not ancestry: the panel is enriched for the derived, out-of-Africa light-skin alleles that archaics (like the earliest-splitting African lineages) simply never carried.

So the one-line takeaway is genes track ancestry; hand-picked SNPs track selection — a caution for anyone reading a pigmentation-SNP panel as if it described who is related to whom. The full logic, figures, and method are in the projection workbook.

Explore the analysis

  • Modern reference PCA (SGDP) — the control that the 159-genome modern reference recovers global human structure before any archaic comparison.
  • The SNP set — how the pigmentation-variant surface was assembled (1,155 SNPs from GWAS, forensic, and gene-network sources) and tagged high- vs low-confidence.
  • Panel coverage & variability — how deeply each genome is sequenced at the panel, read at the correct coordinates, with the genotype and depth heatmaps.
  • Archaic projection — the three-way comparison behind the headline result, worked through step by step.

The approach

SEPIA is a QC-first re-analysis. It continues the honors thesis PAINT (Lily Heald, 2026) but rebuilds the analysis from clearly-traced inputs, because archaic DNA is patchy and shallow and pigmentation genes don’t behave like ancestry markers. Every figure is regenerated from a documented pipeline so its provenance is auditable, and known pitfalls — genome-build mismatches, degenerate PCAs, imputation artefacts — are checked rather than assumed away. The work runs as a phased plan: trace inputs → one verified genome build → positive controls → sample QC → the pigmentation analyses → reproducible infrastructure → writing.

The archaic set is chosen for reliability, not count. SEPIA genotypes the 9 genomes — 5 high-coverage (Altai/Denisova 5, Vindija 33.19, Denisova 3, Chagyrskaya 8, Denisova 25) plus 4 low-coverage (Les Cottés, Denisova 11, Goyet, Mezmaiskaya 1) — that can contribute trustworthy calls at the panel, and deliberately excludes exome-only, redundant, and ultra-low-coverage genomes by documented inclusion criteria.

Status

  • v1 — honors thesis (PAINT, 2026, Lily Heald). Framed the feasibility question: how complete and deep is archaic sequencing at pigmentation loci, and where do archaic samples fall relative to modern humans?
  • In progress — SEPIA. Rebuilding on clearly-traced data: a completed 159-genome modern reference, a harmonised genome build, genotype-likelihood handling for low-coverage genomes, the extended SNP surface, and the redone principal-component projections — plus a Trends in Genetics review.

What the work finds (so far)

Archaic sequencing is uneven and often shallow, and missing data is non-random (older samples have more of it). On their own, ascertained pigmentation SNPs don’t separate populations the way the whole genome does — and where they do deviate reflects selection history, not ancestry. Confident, species-wide claims about archaic pigmentation are not well supported; the contribution is an honest map of what the genetics does and doesn’t show.

Data & provenance

  • Simons Genome Diversity Project (SGDP) — modern human reference genomes.
  • Max Planck Institute for Evolutionary Anthropology (MPI-EVA) — archaic Neanderthal and Denisovan genomes.
  • NHGRI-EBI GWAS Catalog, plus forensic (HIrisPlex-S) and gene-network sources — pigmentation-associated variants.
  • NASA Earth Observations — UV-index climatology.

Per-file provenance — and where the large, cluster-only data live — is documented in data/README.md; per-genome sources are in the provenance tables.

Team

Tina Lasisi (PI / advisor) · Lily Heald · Yemko Pryor.

Code is released under the MIT License; written content and figures under CC BY 4.0 — see License.