Review Papers Claim by Claim

 Quick Answer Copied

 Long Answer

What the paper claims and how it is supported

The preprint defines RAS (Rare Allele Sharing) as an outgroup-F3-like statistic computed only on SNPs ascertained to be rare (derived allele frequency between f_min and f_max in a reference panel), plus RASD difference-statistics analogous to F4, and a linear-regression ancestry decomposition based on RAS profiles against multiple right populations with block-Jackknife errors RAS method definition and ancestry decomposition.

Evidence type 1 — simulations: msprime 3×3 population grids (N_e=20,000; 50 diploids each; 20 chromosomes × 100 Mb) across scaled migration rates 4mN_e = 1–5000. RAS still separates populations at 4mN_e=100 where outgroup-F3 already shows overlap, and even at 4mN_e=2000 RAS retains some power while F3 appears random; a 1.2M-variant "1240K-like" panel is substantially less resolved Simulation benchmark of RAS vs F3. For ancestry decomposition (population 4 = 50/50 mix of populations 0 and 8), RAS-based estimates show lower true error and self-error than F3-based estimates at high migration, with a crossover near 4mN_e≈100 where common-variant statistics slightly win Ancestry decomposition accuracy benchmark.

Evidence type 2 — real data: using the harmonized 1kGP+HGDP dataset (~133 million biallelic SNPs screened; 2184 individuals across 16 European reference populations, supplemented with Danish, Dutch and Swedish allele counts) and African outgroups, RAS amplifies separation of HGDP Russians and Basques toward FIN/IBS references and clarifies isolated groups (Sardinian, Orcadian) relative to F3 on all sites or 1240K European real-data RAS results. On Early Medieval England shotgun genomes, RASD(England_CNE, England_WBI; R1, R2) reaches Z-scores >10 for some reference pairs (NL/SE better than FIN), outperforming F4 on 1240K; RASD-based individual CNE estimates correlate ~0.65–0.90 with the supervised-admixture estimates from Gretzinger et al. (2022), peaking at f_max ≈ 0.6–1% CNE/WBI ancient-Britain RASD results.

Critical assessment, limitations and blind spots

• Sample-size dependence: the authors themselves note rare-allele frequency estimates have larger relative error, so resolution improves directly with sample size; at very low migration RAS is noisier than F3, and at the extreme 4mN_e=5000 the advantage is weak Author-stated limitations of RAS.
• Assumption of no left-to-right gene flow: the ancestry model assumes no gene flow from target/sources into right populations; the authors argue RAS is only sensitive to very recent reverse flow, but this remains an approximation not exhaustively stress-tested No-reverse-gene-flow assumption.
• Missing comparisons: no formal qpWave/qpAdm-style symmetry test or comparison against IBD/ARG-based fine-resolution methods (e.g., those cited for future work) is provided — the authors explicitly defer this to future work.
• Document-quality caveat: the parsed text is a bioRxiv preprint (July 2025) with peer review not yet demonstrated, and the machine-parsed author list is visibly corrupted (citation strings mixed in); the reliably identified authors are Lei Huang, Thiseas C. Lamnidis and corresponding author Stephan Schiffels. Broader rare-variant literature confirms both the promise and the known power/sampling pitfalls of rare-variant analyses Rare variant sequencing design study Biobank rare-variant sampling study.

What would change this conclusion: independent replication on non-European datasets (the empirical validation is Europe-centric), adversarial demographies with reverse gene flow, and head-to-head benchmarking against IBD/ARG methods; failure of RAS to beat F-statistics in such tests would falsify the central claim. Note the frequency-cutoff dependence of Z-scores means results are ascertainment-sensitive rather than parameter-free. All data and code (RAS_tools, RAS_exploration on GitHub) are public, aiding verification.

 BGPT Paper Review

Study Novelty

Formalizes previously informal rare-allele methods (doubleton sharing, low-frequency ancestry ratios) into a coherent F-statistic-like framework with software; builds on established F-statistic theory but is a genuinely new contribution to archaeogenetics tooling.

Scientific Quality

Careful simulation design and real-data validation with public data/code; however, it is an unreviewed preprint, the parsed author metadata is corrupted, empirical validation is Europe-only, and the no-reverse-gene-flow assumption is argued rather than rigorously tested.

Study Generality

The method applies to any biallelic SNP dataset (modern or ancient, any organism) wherever whole-genome data exist, though all empirical demonstrations are human European.

Study Usefulness

Provides immediately usable R software, higher-resolution structure detection, and potentially more accurate ancestry proportions for closely related sources — a real gap in current qpAdm/F-statistic workflows.

Study Reproducibility

All input datasets are public (1kGP+HGDP harmonized, GenomeDenmark, GoNL, SweGen, published ancient British data) and both code repositories are on GitHub with documentation; no proprietary reagents or data.

Explanatory Depth

Derives RAS/RASD from F-statistic theory with explicit algebra, frequency-cutoff interpretations tied to drift/gene-flow timescales, and error decomposition (sampling noise vs jackknife fit error).