LDx: estimation of linkage disequilibrium from high-throughput pooled resequencing data

High-throughput pooled resequencing offers significant potential for whole genome population sequencing. However, its main drawback is the loss of haplotype information. In order to regain some of this information, we present LDx, a computational tool for estimating linkage disequilibrium (LD) from pooled resequencing data. LDx uses an approximate maximum likelihood approach to estimate LD (r2) between pairs of SNPs that can be observed within and among single reads. LDx also reports r2 estimates derived solely from observed genotype counts. We demonstrate that the LDx estimates are highly correlated with r2 estimated from individually resequenced strains. We discuss the performance of LDx using more stringent quality conditions and infer via simulation the degree to which performance can improve based on read depth. Finally we demonstrate two possible uses of LDx with real and simulated pooled resequencing data. First, we use LDx to infer genomewide patterns of decay of LD with physical distance in D. melanogaster population resequencing data. Second, we demonstrate that r2 estimates from LDx are capable of distinguishing alternative demographic models representing plausible demographic histories of D. melanogaster.

💡 Research Summary

The paper introduces LDx, a computational pipeline designed to estimate linkage disequilibrium (LD) from high‑throughput pooled resequencing data, a setting in which individual haplotype information is lost due to the mixing of DNA from many individuals. LDx exploits the fact that a single sequencing read (or a paired‑end fragment) can sometimes span two polymorphic sites, thereby providing direct evidence of the haplotype carried by that fragment. Using an approximate maximum‑likelihood framework, LDx computes the classic r² statistic for each SNP pair that is observed together on at least one read. In addition to this “read‑based” estimate, LDx also reports a conventional r² derived solely from observed genotype frequencies, allowing users to compare the two sources of information and to retain estimates when read‑based data are sparse.

The authors validate LDx by comparing its r² estimates to those obtained from individually sequenced Drosophila melanogaster lines. Across a range of read depths, LDx shows a very high correlation (Pearson r > 0.9) with the gold‑standard individual‑based estimates, especially when the average depth exceeds 30×. They further explore how quality filters—minimum mapping quality, base quality, and read length—affect performance. Stricter filters reduce spurious LD signals and sharpen the decay curve of LD with physical distance.

Through extensive simulations, the paper quantifies the influence of key experimental parameters. Read depth below 10× yields biased r² values, while depths between 20× and 40× produce near‑unbiased estimates with narrow confidence intervals. Read length is equally critical: reads longer than 100 bp dramatically increase the probability that two SNPs fall on the same fragment, whereas short reads (≤50 bp) limit the number of observable SNP pairs and thus the resolution of LD estimates. The authors also show that SNP density and inter‑SNP distance modulate the number of usable read‑based observations, emphasizing the need for paired‑end or longer reads to capture long‑range LD.

Two illustrative applications demonstrate LDx’s utility. First, the authors apply LDx to a natural D. melanogaster population and reconstruct the genome‑wide decay of LD with distance. The resulting curve shows high r² for SNPs within 1 kb and a rapid decline beyond 10 kb, matching patterns previously reported from individual sequencing studies. Second, they use LDx to discriminate between competing demographic scenarios: a recent rapid population expansion versus a long‑term stable population size. Simulated data under each model generate distinct r² distributions, and LDx applied to real pooled data successfully identifies the model that best fits the observed LD pattern, illustrating its potential for demographic inference.

The paper acknowledges limitations. Very low coverage and short reads restrict the number of SNP pairs that can be observed on the same fragment, leading to reduced accuracy. Regions with copy‑number variation or structural rearrangements may introduce mapping artefacts that bias LD estimates. Consequently, the authors recommend experimental designs that achieve at least 20× coverage with reads of 100 bp or longer and that apply stringent quality thresholds.

In summary, LDx bridges the gap between cost‑effective pooled resequencing and the need for haplotype‑level LD information. By delivering reliable r² estimates from read‑based evidence, it enables genome‑wide LD mapping, demographic model testing, and other population‑genetic analyses without the expense of sequencing each individual separately. Future extensions could incorporate longer insert libraries, copy‑number correction, and adaptation to non‑model organisms, further expanding the method’s applicability across evolutionary and conservation genomics.