SEPIA — source-paper references & data provenance (every genome)
SEPIA — source-paper references & data provenance (every genome)
Provenance for every genome in the project, verified against MPI-EVA / ENA / the source papers (2026-07-23). This is the citation + data-source manifest; it feeds the Q1 provenance manifest.
On the PDFs. Actual paper PDFs are downloaded into plan/papers/pdf/ which is git-ignored — the repo is public and most of these are paywalled (Nature/Science/PNAS), so committing them would redistribute copyrighted material. Open-access / author-copy / preprint PDFs were pulled automatically (marked ✅ local); paywalled ones are linked for download via institutional (UMich) access into the same folder (marked 🔒).
Two clarifications that matter for the reanalysis:
- The MPI-EVA high-coverage VCFs are damage-aware
snpADgenotype calls, not statistically phased haplotypes. "Phased VCF" = a downstream step we'd run ourselves if needed. QC details per release:VCF_QC.md. - Vindija 87 is the same individual as Vindija 33.19 (the 30× genome) — so called genotypes for it exist via the Vindija 33.19 VCF; the 1.3× "Vi87" BAM is a separate low-coverage library of the same person.
Tier A — high-coverage archaics: use the published snpAD genotype VCFs (don't re-call)
| Genome | Site · type | Coverage | Paper | DOI | Data (VCF / BAM) | |
|---|---|---|---|---|---|---|
| Altai Neanderthal (= Denisova 5) | Denisova Cave · Neanderthal | ~52× | Prüfer et al. 2014, Nature 505:43–49 | 10.1038/nature12886 | VCF (snpAD 2017 reprocess): cdna.eva.mpg.de/neandertal/Vindija/VCF/Altai/chr{1..22,X}_mq25_mapab100.vcf.gz; original: cdna.eva.mpg.de/neandertal/altai/AltaiNeandertal/VCF/; BAM: .../AltaiNeandertal/bam/ |
🔒 Nature |
| Vindija 33.19 | Vindija Cave, Croatia · Neanderthal | ~30× | Prüfer et al. 2017, Science 358:655–658 | 10.1126/science.aao1887 | VCF: cdna.eva.mpg.de/neandertal/Vindija/VCF/Vindija33.19/chr{1..22,X}_mq25_mapab100.vcf.gz; ENA PRJEB21157; BAM under ftp.eva.mpg.de/neandertal/Vindija/ |
✅ pruefer2017_vindija33_science.pdf |
| Chagyrskaya 8 | Chagyrskaya Cave · Neanderthal | ~27–28× | Mafessoni et al. 2020, PNAS 117:15132–15136 | 10.1073/pnas.2004944117 | VCF: ftp.eva.mpg.de/neandertal/Chagyrskaya/VCF/chr{1..22,X}.noRB.vcf.gz; mask: .../Chagyrskaya/FilterBed/; BAM: .../Chagyrskaya/BAM/ |
🔒 PNAS (PMC likely OA) |
| Denisova 3 | Denisova Cave · Denisovan | ~30× | Meyer et al. 2012, Science 338:222–226 | 10.1126/science.1224344 | VCF (snpAD 2017): cdna.eva.mpg.de/neandertal/Vindija/VCF/Denisova/chr{1..22,X}_mq25_mapab100.vcf.gz; original: cdna.eva.mpg.de/denisova/VCF/human/; BAM: cdna.eva.mpg.de/denisova/BAM/ |
🔒 Science |
| Denisova 25 ⚠ preprint | Denisova Cave (S. Chamber, ~200 ka molar) · Denisovan | ~24× | Peyrégne et al. 2025, bioRxiv | 10.1101/2025.10.20.683404 | VCF: cdna.eva.mpg.de/denisova/Den25/VCF/chr{1..22,X,Y}.Den25.L35MQ25.B30.map35_100.vcf.gz; mask: .../Den25/FilterBed/; BAM: .../Den25/BAM/ |
✅ den25_peyregne2025_biorxiv.pdf |
The Altai + Vindija 33.19 + Denisova 3 VCFs share one 2017 snpAD release (same pipeline, reference, filters, one README) → most directly comparable, and the Mez1 low-cov snpAD calls sit in the same release tree. The original Altai (2014) and Denisova 3 (2012) releases are GATK UnifiedGenotyper calls — prefer the snpAD reprocess for a harmonized cohort. Chagyrskaya 8 and Denisova 25 are separate snpAD releases with their own masks — apply each release's FilterBed and harmonize before merging (see VCF_QC.md). Denisova 25 is a preprint whose data are not yet deposited → provisional. The published VCFs are not the final call set — the companion FilterBed mask (cov≥10, GC central-95%, tandem-repeats, indels) must be applied on top.
Tier B — low-coverage archaics: BAM only → ANGSD genotype likelihoods (no published VCF)
| Genome | Site · type | Coverage | Paper | DOI | Archive | |
|---|---|---|---|---|---|---|
| Denisova 11 "Denny" | Denisova Cave · F1 hybrid (Nea mother × Den father) | ~2.6× | Slon et al. 2018, Nature 561:113–116 | 10.1038/s41586-018-0455-x | ENA PRJEB24663 (raw); BAM | 🔒 Nature |
| Goyet Q56-1 | Goyet, Belgium · late Neanderthal | 2.2× | Hajdinjak et al. 2018, Nature 555:652–656 | 10.1038/nature26151 | ENA PRJEB21870; EVA neandertal/GoyetQ56-1/ |
🔒 (OA: PMC6485383) |
| Les Cottés Z4-1514 | Les Cottés, France · late Neanderthal | 2.7× | Hajdinjak et al. 2018 | 10.1038/nature26151 | ENA PRJEB21875; EVA LesCottes_Z4-1514/ |
🔒 (OA: PMC6485383) |
| Spy 94a | Spy, Belgium · late Neanderthal | 1.0× | Hajdinjak et al. 2018 | 10.1038/nature26151 | ENA PRJEB21883; EVA Spy94a/ |
🔒 (OA: PMC6485383) |
| Mezmaiskaya 2 | Mezmaiskaya, Russia · late Neanderthal (~44 ka) | 1.7× | Hajdinjak et al. 2018 | 10.1038/nature26151 | ENA PRJEB21881; EVA Mezmaiskaya2/ |
🔒 (OA: PMC6485383) |
| Vindija 87 (= Vi 33.19 individual) | Vindija, Croatia · late Neanderthal | 1.3× | Hajdinjak et al. 2018 | 10.1038/nature26151 | ENA PRJEB21882; EVA Vindija87/. Called genotypes: use the Vi 33.19 VCF (Tier A). |
🔒 (OA: PMC6485383) |
| Mezmaiskaya 1 | Mezmaiskaya, Russia · Neanderthal (~65 ka) | ~0.5× + 1.4× ≈ 1.9× | Prüfer et al. 2017 (added); Prüfer 2014 (orig.) | 10.1126/science.aao1887 | ENA PRJEB21195; a low-cov snpAD VCF EXISTS: cdna.eva.mpg.de/neandertal/Vindija/VCF/Mez1/ (+ BAM Mezmaiskaya1/) |
✅ (Prüfer 2017 pdf) |
| Hohlenstein-Stadel | Hohlenstein-Stadel, Germany · early Neanderthal (~120 ka) | ~0.05× (168 Mbp) | Peyrégne et al. 2019, Sci. Adv. 5:eaaw5873 | 10.1126/sciadv.aaw5873 | ENA PRJEB29475; EVA Hohlenstein-Stadel/BAM/All_sequences/L5386.bam |
✅ peyregne2019_hst_scladina_sciadv.pdf |
| Scladina I-4A | Scladina, Belgium · early Neanderthal (~120 ka) | ~0.02–0.03× (78 Mbp) | Peyrégne et al. 2019 | 10.1126/sciadv.aaw5873 | ENA PRJEB29475 (2° ERP111781); run ERR9741105 = G2160.bam; EVA Scladina/BAM/ |
✅ (Peyrégne 2019 pdf) |
Tier C — no usable genome-wide data (excluded → 14 analyzed, not 15)
| Genome | Status | Paper | DOI | Note |
|---|---|---|---|---|
| El Sidrón 1253 | exome-capture only (~12.5× over coding regions); no shotgun whole-genome | Castellano et al. 2014, PNAS 111:6666–6671 | 10.1073/pnas.1405138111 | EVA neandertal/exomes/ (BAM+VCF, coding only). Not comparable to the genome-wide panel → correctly excluded. |
Modern reference & methods
| Item | Paper | DOI | Source |
|---|---|---|---|
| SGDP (15 moderns used) | Mallick et al. 2016, Nature 538:201–206 | 10.1038/nature18964 | Seven Bridges CGC, remapped to hs37d5/hg19 (sgdp_merge LP-IDs recorded in the Q1 manifest) |
| snpAD genotyper (the high-cov VCF caller) | Prüfer 2018, Bioinformatics 34:4165–4171 | 10.1093/bioinformatics/bty507 | damage-aware ancient-DNA genotyper |
| Reference genome | hs37d5 / GRCh37 (hg19) | — | 1000 Genomes decoy build; all archaic BAMs and SGDP are aligned to it |
Flagged / unconfirmed
- Denisova 25 — preprint (not yet peer-reviewed); EVA
Den25/files are live but formal ENA/SRA accession not yet stated; treat as provisional. - El Sidrón exome ENA accession — unconfirmed; authoritative access is the EVA
exomes/URL. - HST / Scladina fold-coverage — derived from the paper's Mbp figures (168 / 78 Mbp ÷ ~3.2 Gb), not stated as fold in the text.
- Chagyrskaya
.noRBfilename meaning — to confirm from the release README (likely "RepeatBed/reliable-blocks mask not yet applied" → applyFilterBed/yourself).
PDFs downloaded locally: den25_peyregne2025_biorxiv.pdf, pruefer2017_vindija33_science.pdf, peyregne2019_hst_scladina_sciadv.pdf. Paywalled papers (Prüfer 2014, Meyer 2012, Slon 2018, Mafessoni 2020, Hajdinjak 2018, Castellano 2014, Mallick 2016) → download via UMich library into plan/papers/pdf/ using the DOIs above.