Adaptive Sentinel Sampling for Founder-Enriched Populations: Quantifying the Resolution Gap in National Genome Programmes and Implications for Genomic Equity

Background: India harbours one of the world's most diverse mosaic of founder-enriched, endogamous communities, each shaped by centuries of genetic isolation and severe demographic bottlenecks. The Genome India Project (GIP), a landmark national genome programme, incorporates broadly similar per-population sampling targets across multiple strata. We hypothesised that this architecture creates a statistically predictable "Resolution Gap" — a measurable probability of missing rare, clinically important private alleles in tribal cohorts — with direct implications for diagnostic equity and precision medicine delivery to historically underrepresented communities. Results: Mathematical audit using the binomial non-detection formula P₀ = (1 − p)2n revealed that the GIP tribal sampling stratum (n = 75) carries a 22.17% non-detection probability (P₀) at 1% minor allele frequency (MAF), rising to 47.15% at 0.5% MAF — meaning that under the binomial sampling model, a variant with true MAF of 0.5% has a 47.15% probability of not being observed in a sample of 75 diploid individuals. Monte Carlo simulation (1,000,000 iterations) confirmed all values to within 0.05%. The proposed Two-Stage Adaptive Sentinel Model (TASM) — triggered by a composite FST > 0.02 and private allele detection ≥2.5% in a sentinel cohort (n = 20) — reduces P₀ to 0.24% at 1% MAF with a modelled total of 8,660 genomes, representing a modelled reduction in total sequencing burden compared to the current uniform-sampling requirement of 10,074 genomes. Conclusions: Under-powered reference databases in founder-enriched populations may contribute to a proposed 'VUS Inflation Trap': absence of population-specific frequency data deprives clinicians of PM2 evidence context under ACMG/AMP guidelines, complicating variant interpretation. Empirical annotation of 2,269 biallelic BRCA1 variants across South Asian subpopulations found that 33.1% had AF=0.0 in gnomAD v2.1.1 and 99.7% lacked a definitive ClinVar classification or entry — consistent with this mechanism operating in real population genomic data, though formal clinical validation through case series in affected populations remains an important future direction. Proof-of-concept evaluation using 1000 Genomes Phase 3 data across BRCA1, BRCA2, and TP53 in five South Asian subpopulations provides preliminary evidence that the trigger architecture identifies population-specific enrichment signals in real genomic data. The framework is potentially applicable to other national genome programmes with heterogeneous population structure, subject to programme-specific empirical calibration. Keywords: adaptive sampling; founder populations; genomic equity; rare variant detection; population genomics; Genome India Project; VUS Inflation Trap; tribal genomics; pangenome representation; non-detection probability