Accounting for long-range correlations in genome-wide simulations of large cohorts

Abstract

Coalescent simulations are widely used to examine the effects of evolution and demographic history on the genetic makeup of populations. Thanks to recent progress in algorithms and data structures, simulators such as the widely-used msprime now provide genome-wide simulations for millions of individuals. However, this software relies on classic coalescent theory and its assumptions that sample sizes are small and that the region being simulated is short. Here we show that coalescent simulations of long regions of the genome exhibit large biases in identity-by-descent (IBD), long-range linkage disequilibrium (LD), and ancestry patterns, particularly when the sample size is large. We present a Wright-Fisher extension to msprime, and show that it produces more realistic distributions of IBD, LD, and ancestry proportions, while also addressing more subtle biases of the coalescent. Further, these extensions are more computationally efficient than stateof-the-art coalescent simulations when simulating long regions, including whole-genome data. For shorter regions, efficiency can be maintained via a hybrid model which simulates the recent past under the Wright-Fisher model and uses coalescent simulations in the distant past.

Inclusion of variants discovered from diverse populations improves polygenic risk score transferability

“…First, while coalescent simulations allow for decreased computational burden, model assumptions may result in inaccurate long-range LD, especially for whole-genome simulations. 30 However, given we only simulated chromosome 20, biases are expected to be modest. 30 We also use a case-control framework for our simulation; therefore, power and potential differences in population PRS accuracy may be even higher if a quantitative trait was used.…”

“… 30 However, given we only simulated chromosome 20, biases are expected to be modest. 30 We also use a case-control framework for our simulation; therefore, power and potential differences in population PRS accuracy may be even higher if a quantitative trait was used. In addition, our simulations assume random mating among admixed individuals and therefore do not reflect the more complex assortative mating that may be observed, which may impact the distribution of local ancestry tract lengths in our simulation and therefore hinder the improvement of PRS accuracy by local ancestry weighting.…”

Sex-biased admixture and assortative mating shape genetic variation and influence demographic inference in admixed Cabo Verdeans

“…After running Refined IBD, we used its gap-filling utility to remove gaps between segments less than 0.6 cM that had at most one discordant homozygote. We then filtered out IBD segments smaller than 5 cM, as short segments below this threshold are difficult to accurately detect 59 . To visualize IBD sharing in We called ROH using GARLIC v.1.1.6 61 , which implements the ROH calling pipeline of Pemberton et al 201229 .…”

GSpace: an exact coalescence simulator of recombining genomes under isolation by distance

“…Rousset, 2000), but cannot be simulated when considering extensions of Kingman's (1982) n-coalescent which assume large sub-population sizes and small migration rates. To simulate small population sizes and high dispersal rates without biases (Nelson et al, 2020), coalescence probabilities exact for small population size (Fu, 2006) must be used in a gen-bygen simulation until the common ancestors of the whole simulated sample have been found. Such simulations are required to assess any inference framework which might for example allow separate estimation of sub-population size, mutation and migration probabilities, something that is not possible under n-coalescent approximations.…”

“…As gen-by-gen algorithms are expected to be slower than those involving n-coalescent approximations, we performed simulations to check the feasibility of simulating genomic data by such algorithms, and compared computation times with those of alternative software based on n-coalescent approximations, such as msprime (Kelleher et al, 2016), FastSimcoal2 (Excoffier et al, 2013), exact coalescence algorithms implemented as DTWF in msprime python package [back-in-time Wright-Fisher simulator, Nelson et al (2020)], IBDSim (Leblois et al, 2009) and forward algorithms, such as SimBit (Matthey-Doret, 2020).…”