Inferring Strain Mixture within Clinical Plasmodium falciparum Isolates from Genomic Sequence Data

We present a rigorous statistical model that infers the structure of P. falciparum mixtures—including the number of strains present, their proportion within the samples, and the amount of unexplained mixture—using whole genome sequence (WGS) data. Applied to simulation data, artificial laboratory mixtures, and field samples, the model provides reasonable inference with as few as 10 reads or 50 SNPs and works efficiently even with much larger data sets. Source code and example data for the model are provided in an open source fashion. We discuss the possible uses of this model as a window into within-host selection for clinical and epidemiological studies.Author Summary: Since the 1960’s researchers have observed that Plasmodium falciparum infections, the cause of most severe malaria, are frequently composed of several different strains of the parasite. In this work, the authors use Bayesian methods on whole genome sequence data to model the structure of these mixtures. Results from this method are consistent with previous approaches but provide finer resolution of these mixtures. As whole genome data in malaria research becomes increasingly common, this work will allow researchers to delve further into the within-host dynamics of the parasite, a much-discussed but previously difficult-to-access aspect of this disease.