An Atlas of Linkage Disequilibrium Across Species

Linkage disequilibrium (LD) is a key metric that characterizes populations in flux. To reach a genomic scale LD illustration, which has a substantial computational cost of O(nm2), we introduce a framework with two novel algorithms for LD estimation: X-LD, with a time complexity of O(n2m) suitable for small sample sizes (n < 104); X-LDR, a stochastic algorithm with a time complexity of O(nmB) for biobank-scale data (B iterations); n the sample size, and m the number of SNPs. These methods can refine the entire genome into high-resolution LD grids, such as more than 9 million grids for UK Biobank samples (approximately 4.2 million SNPs). The efficient resolution for genome-wide LD leads to intriguing biological discoveries. I) High-resolution LD illustrations revealed how the pericentromeric regions and the HLA region lead to intense and extended LD patterns. II) Two universal LD patterns, identified as Norm I and Norm II patterns, provide insights on the evolutionary history of populations and can also highlight genomic regions of deviation, such as chromosomes 6 and 11 or ncRNA regions. III) The results of our innovative LD decay method aligned with the LD decay scores of 59.5 for Europeans, 60.2 for East Asians, and 33.2 for Africans; correspondingly, the length of the LD was approximately 2.85 Mb, 2.18 Mb, and 1.58 Mb for these three ethnicities. Rare or imputed variants universally increased LD. IV) An unprecedented LD atlas for 25 reference populations contoured interspecies diversity in terms of their Norm I and Norm II LD patterns, highlighting the impact of refined population structure, quality of reference genomes, and uncovered a profound status quo of these populations. The algorithms have been implemented in C++ and are freely available (https://github.com/gc5k/gear2).