Review Papers Claim by Claim

Concise appraisal

The study assembles ~8,000 modern and ancient South Asian genomes and shows that the Eurasian LP variant -13910C>T (paper uses -13.910:C>T notation) is geographically clinal across South Asia and that genome-wide Steppe ancestry explains most variation in nonpastoralists, while two pastoralist groups (Toda and Gujjar) show strong, recent positive selection on Steppe-derived LCT haplotypes (selection coefficients s approx 0.03–0.05) Moorjani et al Preprint 2025

Paper Review Revisited Lactase Persistence in South Asia

High level summary

The authors combine 7,962 present-day whole genomes and 129 ancient Swat Valley genomes (total ~8,091 samples) to map the distribution and evolutionary history of the Eurasian LP variant -13.910:C>T across South Asia, and to separate the roles of ancestry, drift, founder events, and selection. They report that (1) -13.910:T is widespread with north–south and east–west clines, (2) Steppe pastoralist ancestry explains nearly all allele frequency variation in nonpastoralist groups (strong correlation r2~0.97), and (3) two pastoralist groups (Toda and Gujjar) are extreme outliers with very high -13.910:T frequencies driven by recent, strong positive selection on Steppe-derived haplotypes with estimated s~0.03–0.05 Moorjani et al Preprint 2025

Detailed strengths

• Exceptional sample breadth: ~8,000 genomes spanning India Pakistan Bangladesh plus temporally deep ancient Swat Valley samples provides statistical power to detect clines and compare ancient vs modern allele trajectories Dataset scope
• Integrated methods: combined PCA ADMIXTURE qpAdm local ancestry IBD haplotype analyses and demographic simulations (msprime) — a modern, complementary toolkit that reduces single-method biases Methods summary
• Conservative sensitivity checks: authors acknowledge and test ancient imputation limitations and perform read-count based calls for ancient samples where possible, and run demographic simulations to test drift-only scenarios for pastoralist high frequencies Ancient data caveats and checks

Main weaknesses and blindspots

1. Dependence on imputed low coverage ancient data at LCT — authors use GLIMPSE-imputed 1240K AADR data; 1240K lacks dense coverage in regulatory LCT enhancers and imputation in complex LD regions can be biased, possibly affecting timing estimates of allele arrival Ancient imputation caution
2. Geographic sampling bias — ancient data concentrated in Swat Valley limits ability to date arrival across wider South Asia; additional ancient samples from Indus and peninsular India would strengthen temporal claims Ancient sampling limitation
3. Assumption that Steppe ancestry introduced -13.910:T — authors present strong haplotype similarity to Steppe pastoralist T haplotypes, but independent emergence or earlier low-frequency presence in unsampled sources cannot be fully excluded without denser ancient coverage and targeted sequencing of LCT region in ancient remains Haplotype origin evidence
4. Potential confounding cultural/dietary data missing — the study notes the need for archaeological/ethnographic measures of raw versus fermented milk consumption; without these, connecting selection intensity to milk consumption remains hypothetical Need for cultural context
5. Winner's curse on selection coefficient estimation — selection s estimates for small endogamous groups with strong founder events (especially Toda) can be upwardly biased; authors use simulations but credible intervals and alternative dominance models should be emphasized more transparently Selection estimation caveat

Key quantitative results (paper derived)

• Total samples analyzed approx 8 091 including 7 962 present day 129 ancient; 45 endogamous groups including pastoralists Toda Gujjar Sample sizes
• Geographic allele frequency gradient: -13.910:T >25% in North India Pakistan (LP ~44.5%) dropping to <10% East India and <5% South India; many South and East endogamous groups lack the allele entirely Allele frequency cline
• Ancient trend: absent in Bronze/Iron Age samples rising in Historic (~9.1%) and Medieval (~25%) periods in Swat Valley samples consistent with later arrival via Steppe gene flow Ancient allele trajectory
• Ancestry correlation: Steppe ancestry explains nearly all variation in -13.910:T among nonpastoralists (reported r2 approx 0.97 p<1e-6) while SAHG ancestry correlates negatively and Iranian farmer ancestry shows weaker positive correlation (r2~0.69) Ancestry correlations
• Pastoralist outliers: Toda 61% derived allele (LP approx 89%); Gujjar 67.5% derived (LP approx 90%); local ancestry deviation test Z-scores indicate significant Steppe enrichment around LCT (Toda Z=8.74 Gujjar Z=4.5) Pastoralist outliers
• Haplotype evidence: derived T haplotypes in South Asia nearly identical to Steppe T haplotypes with average differences ~0.19 across core SNPs and markedly lower diversity relative to C haplotypes consistent with recent introduction and selection Haplotype similarity
• Selection coefficients: estimated s Toda approx 0.047 (s dominant 0.035), Gujjar s approx 0.033 (s dominant 0.031) compared with European estimates ~0.018 indicating strong recent selection if estimates hold Selection coefficient estimates

Visual reproduction of key figure

Note: regional LP percentages derived from paper reported allele frequency to LP conversions (dominant model) and are approximate: North India Pakistan LP~44.5% East India LP~13.77% South India LP~6.56% Regional LP estimates

Interpretation and alternative explanations

The authors' primary interpretation is that in most South Asians, the -13.910:T frequency pattern is explained by Steppe-derived admixture introduced ~3500 years ago and that only pastoralists like Toda and Gujjar show additional recent positive selection boosting local frequencies. This interpretation is robustly supported by (i) high haplotype identity to Steppe T haplotypes (low divergence), (ii) genome-wide correlation between Steppe ancestry and allele frequency (r2~0.97), (iii) local ancestry deviation tests and elevated IBD around LCT in pastoralists, and (iv) demographic simulations showing drift alone is unlikely to reach observed frequencies in those pastoralists Evidence weight

However the alternative hypotheses that merit further testing include:

• Earlier low frequency presence of the allele from un-sampled Near Eastern sources that was later boosted by Steppe admixture and selection — resolving this needs denser ancient coverage across the Indus region and peninsular India.
• Selection driven by immunological or pleiotropic effects at the wider 2q21.3 locus (e.g., UBXN4 DARS1 effects) rather than lactose digestion per se — the authors mention recent reports that LCT region affects immune genes which would alter the selection narrative and require functional/phenotypic data to adjudicate Pleiotropy possibility

How convincing is the selection claim in Toda and Gujjar?

Evidence in favor:

• Strong local enrichment of Steppe ancestry at the LCT region far exceeding genome-wide expectations (LAD Z 8.74 and 4.5) LAD evidence
• Enriched IBD sharing and unusually long, low diversity Steppe-derived haplotypes consistent with recent selective sweep IBD and haplotypes
• Coalescent simulations indicate drift and founder effects are unlikely to produce observed high frequencies under the inferred demographic histories of these populations Simulation support

Limitations that temper conviction:

• Small sample sizes per endogamous group (n 18 20) increase variance and the possibility that sampling and recent relatedness inflate frequency estimates despite removal of close relatives — authors took care to remove first-degree relatives but small N remains a concern Small group N caveat
• High founder intensity in Toda (estimated intensity 8.5 with bottleneck ~532 years ago) complicates selection inference because bottlenecks can mimic sweeps; although simulations argue drift unlikely to fully explain the result they do not eliminate some joint bottleneck plus selection complexities Founder event complexity

Recommended follow ups to strengthen and falsify claims

1. Targeted high depth sequencing of the LCT regulatory region in ancient samples across the Indus and peninsular India to directly genotype causal variants without imputation and to provide finer temporal resolution of allele arrival.
2. Phenotype surveys (lactose tolerance tests) and ethnographic dietary quantification of milk consumption in Toda Gujjar and neighboring nonpastoralists to link genotype selection pressure to actual raw milk intake versus fermented dairy consumption.
3. Functional assays and eQTL work in relevant intestinal and immune cell types to test whether -13.910:T or linked haplotypes modulate expression of LCT or neighboring immune-related genes (UBXN4 DARS1) — to evaluate pleiotropy hypotheses.
4. Expanded simulations integrating joint bottleneck plus various dominance and timing models with explicit uncertainty quantification (credible intervals) to quantify robustness of s estimates under realistic pedigree structures.

Confidence and net assessment

Evidence strength for the paper's major claims is high for the ancestry-driven explanation across most South Asians (multiple methods converging; r2~0.97) and moderate-to-high for selection in Toda and Gujjar given local ancestry enrichment, IBD, and simulations — but caveats remain because of small group sizes, founder events, and limited ancient spatial coverage Overall assessment and caveats

Actionable items for researchers

• Obtain targeted high coverage ancient LCT region sequences from Indus and peninsular contexts.
• Collect standardized phenotypic lactose tolerance and milk consumption data from Toda Gujjar and comparative groups.
• Run locus specific expression and chromatin assays (MPRA, luciferase) for -13.910 and linked variants in intestinal cell models.

Reproducibility and data/code availability

The authors used public resources and list data accessions (LASI DAD via NIAGADS EGAS00001002921 GenomeAsia browser and SRA BioProjects) and used standard open methods (GLIMPSE SHAPEIT4 msprime hap-IBD) — these choices support reproducibility provided ancient imputation parameters and exact code paths are released (the authors link to MoorjaniLab GitHub for references) Data availability

If you want I can run the bioinformatics agent to reproduce the LCT local ancestry plots I used here and run additional robustness simulations for selection coefficient uncertainty.