The Ghost in the Genome: How Population Structure Distorts Ancient DNA Models of Indian Ancestry

In statistical genetics, linkage disequilibrium (LD)—the non-random association of alleles at different positions across a chromosome—serves as a primary clock. When two previously separated populations meet and intermarry, their distinct genetic markers enter the newly formed population together. Over subsequent generations, genetic recombination acts like a pair of biological scissors, snipping these inherited blocks of DNA into smaller and smaller segments. By measuring the length of these shared segments, geneticists can estimate when an ancestral admixture event took place.

However, this statistical clock rests on an underlying assumption: that mating within each reference group occurs randomly. When endogamy—the custom of marrying strictly within a specific group or subgroup—is present, this assumption breaks down. Non-random mating preserves co-occurring genetic variants across distinct genetic loci, creating statistical correlations that resemble physical genetic linkage. A new methodological preprint published on bioRxiv (Hui and Burt, 2026) addresses this exact issue, demonstrating how population structure and structured mating create false signals of linkage disequilibrium and offering a mathematical framework to recalibrate genetic models.

This statistical distortion is particularly relevant to the genetic history of South Asia. Over the past decade, ancient DNA studies have modeled the subcontinent's genetic history through various mixing events involving three primary ancestral streams: Ancestral South Indian (ASI), Ancestral North Indian (ANI), and Austroasiatic or Tibeto-Burman lineages. A central landmark in this literature—the comprehensive study led by Vagheesh Narasimhan and David Reich—established that major admixture between northwestern and indigenous peninsular groups took place primarily between 2000 BCE and 100 BCE. Following this period, the genetic data indicated a widespread transition toward strict endogamy, fixing these genetic signatures within isolated subgroups.

The mathematical problem introduced by uncorrected endogamy is straightforward. When individuals marry exclusively within a restricted subgroup over dozens of generations, specific combinations of genetic variants are passed down together repeatedly, even if they sit far apart on a chromosome or on entirely different chromosomes. Standard computational tools interpret these correlated variants as long, unbroken ancestral tracts. Because long ancestral tracts are the hallmark of recent admixture, standard algorithms misread endogamy-driven correlations as evidence of recent population mixing.

When applied to ancient DNA datasets, this bias can distort timeline calculations. In datasets where population structure is ignored, admixture dates can appear systematically more recent than they actually were. Alternatively, lingering correlation between markers can obscure the true age at which isolated subgroups stopped intermarrying.

The methodology detailed in the new bioRxiv paper introduces an explicit correction for non-random mating into linkage disequilibrium calculations (Hui and Burt, 2026). By testing for population structure before evaluating allele correlations, the mathematical model isolates real physical linkage—caused by shared ancestral chromosomal blocks—from fake linkage induced by structured endogamy (Hui and Burt, 2026).

For South Asian paleogenomics, applying this structural adjustment changes how data from ancient skeletal remains are interpreted. Ancient genomes recovered from sites like Rakhigarhi, analyzed by Vasant Shinde and colleagues, demonstrate a baseline population structure in Bronze Age South Asia that lacked the Steppe-related ancestry that arrived later. When geneticists attempt to trace how and when Steppe-related pastoralist ancestry mixed with local Indus Valley Horizon populations, they rely on accurate decay rates of linkage disequilibrium.

Evidence shows that accounting for endogamy-induced correlation shifts the estimated dates of initial admixture. Rather than representing a uniform, rapid wave of mixing across the subcontinent, corrected models indicate a more protracted process. Local population pockets preserved distinct genetic structures alongside neighboring groups for centuries, generating localized background correlations that inflated previous admixture calculations.

Furthermore, the recalibrated model refines the timeline for when endogamy became the dominant social structure in South Asia. Earlier genetic estimates dated the broad shift from pan-regional mixing to strict endogamy at roughly 2,000 years ago. By stripping out the confounding correlation introduced by micro-population structures, updated statistical tools can differentiate between genuine ongoing admixture and the long-term preservation of ancestral alleles inside closed genetic pools.

In population genetics, a model is only as accurate as its underlying statistical assumptions. The identification of structure-induced linkage disequilibrium does not invalidate the broad three-stream framework of Indian ancestry. Instead, it provides geneticists with a more precise tool, replacing simplified assumptions about random mating with mathematical corrections that reflect actual historical population structures (Hui and Burt, 2026).