Complex genetic admixture histories reconstructed with Approximate Bayesian Computations

Admixture is a fundamental evolutionary process that has influenced genetic patterns in numerous species. Maximum-likelihood approaches based on allele frequencies and linkage-disequilibrium have been extensively used to infer admixture processes from dense genome-wide datasets mostly in human populations. Nevertheless, complex admixture histories, beyond one or two pulses of admixture, remain methodologically challenging to reconstruct, especially when large datasets are unavailable. We develop an Approximate Bayesian Computations (ABC) framework to reconstruct complex admixture histories from independent genetic markers. We built the software package to simulate independent SNPs in a two-way admixed population for scenarios with multiple admixture pulses, or monotonically decreasing or increasing admixture at each generation; drawing model-parameter values from prior distributions set by the user. For each simulated dataset, we calculate 24 summary statistics describing genetic diversity and moments of individual admixture fraction. We coupled with existing ABC algorithms and investigate the admixture history of an African American and a Barbadian population. Results show that Random-Forest ABC scenario-choice, followed by Neural-Network ABC posterior parameter estimation, can distinguish most complex admixture scenarios and provide accurate model-parameter estimations. For both admixed populations, we find that monotonically decreasing contributions over time, from the European and African sources, explain the observed data more accurately than multiple admixture pulses. Furthermore, we find contrasted trajectories of introgression decay from the European and African sources between the two admixed populations. This approach will allow for reconstructing detailed admixture histories in numerous populations and species, particularly when maximum-likelihood methods are intractable.