Recent Demographic History Inferred by High-Resolution Analysis of Linkage Disequilibrium

TLDR

Inferring recent changes in effective population size is crucial for conservation and human history, yet current methods cannot capture complex recent demographic trajectories. The authors aim to develop a framework that infers a population’s demographic history over the past 100 generations from the linkage‑disequilibrium spectrum of contemporary samples. Their model incorporates cumulative contributions from all prior generations and uses a genetic algorithm to fit a sequence of historical Ne values, applicable to samples ranging from large cohorts to fewer than ten individuals and to various genotyping formats. Simulations demonstrate the method’s robustness to genotyping errors, sample heterogeneity, admixture, and subpopulation structure, outperforming leading approaches for recent timeframes, and real‑data analyses in humans and animals corroborate previous estimates and historical records.

Abstract

Abstract Inferring changes in effective population size (Ne) in the recent past is of special interest for conservation of endangered species and for human history research. Current methods for estimating the very recent historical Ne are unable to detect complex demographic trajectories involving multiple episodes of bottlenecks, drops, and expansions. We develop a theoretical and computational framework to infer the demographic history of a population within the past 100 generations from the observed spectrum of linkage disequilibrium (LD) of pairs of loci over a wide range of recombination rates in a sample of contemporary individuals. The cumulative contributions of all of the previous generations to the observed LD are included in our model, and a genetic algorithm is used to search for the sequence of historical Ne values that best explains the observed LD spectrum. The method can be applied from large samples to samples of fewer than ten individuals using a variety of genotyping and DNA sequencing data: haploid, diploid with phased or unphased genotypes and pseudohaploid data from low-coverage sequencing. The method was tested by computer simulation for sensitivity to genotyping errors, temporal heterogeneity of samples, population admixture, and structural division into subpopulations, showing high tolerance to deviations from the assumptions of the model. Computer simulations also show that the proposed method outperforms other leading approaches when the inference concerns recent timeframes. Analysis of data from a variety of human and animal populations gave results in agreement with previous estimations by other methods or with records of historical events.