Approaches to estimating inbreeding coefficients in clinical isolates of Plasmodium falciparum from genomic sequence...

The advent of whole-genome sequencing has generated increased interest in modelling the structure of strain mixture within clinical infections of Plasmodium falciparum The life cycle of the parasite implies that the mixture of multiple strains within an infected individual is related to the out-crossing rate across populations, making methods for measuring this process in situ central to understanding the genetic epidemiology of the disease.

This paper derives a set of new estimators for inferring inbreeding coefficients using whole genome sequence read count data from P. falciparum clinical samples, which provides resources to assess within-sample mixture that connect to extensive literatures in population genetics and conservation ecology. Features of the P. falciparum genome mean that standard methods for inbreeding coefficients and related F-statistics cannot be used directly. After reviewing an initial effort to estimate the inbreeding coefficient within clinical isolates of P. falciparum, several generalizations using both frequentist and Bayesian approaches are provided. A simpler, more intuitive frequentist estimator is shown to have nearly identical properties to the initial estimator both in simulation and in real data sets. The Bayesian approach connects these estimates to the Balding-Nichols model, a mainstay within genetic epidemiology, and a possible framework for more complex modelling. A simulation study shows strong performance for all estimators with as few as ten variants. Application to samples from the PF3K data set indicate significant across-country variation in within-sample mixture. Finally, a comparison with results from a recent mixture model for within-sample strain mixture show that inbreeding coefficients provide a strong proxy for these more complex models.

This paper provides a set of methods for estimating inbreeding coefficients within P. falciparum samples from whole-genome sequence data, supported by simulation studies and empirical examples. It includes a substantially simple estimator with similar statistical properties to the estimator in current use. These methods will also be applicable to other species with similar life-cycles. Implementations of the methods described are available in an open-source R package pfmix. Estimates for the PF3K public data release are provide as part of this resource.