Rare variation in malaria parasites biases population-genetic inference

Abstract

Understanding how pathogens evolve is fundamental to disease control and is a basic question in evolutionary biology, yet pathogens with complex life cycles violate assumptions of classic evolutionary models. Genetic analyses of the malaria parasite Plasmodium falciparum have shown multiple surprising patterns. For example, empirical analyses produce effective population size estimates that vary by orders of magnitude depending on the method and an excess of genes with elevated nonsynonymous variation (measured as{pi} N /{pi}S). Here, reanalyzing genomic data from 18 worldwide populations, I show that these observations directly follow from distributions of genetic variation enriched for rare variants. Multiple potential mechanisms may increase the proportion of rare variants, including host expansions, lifecycle dynamics, population structure, or selection. The observed genealogies are more consistent with a multiple-merger coalescent than a Kingman coalescent, causing common summary statistics to be biased in predictable directions. In particular, the abundance of rare variants interacts with the mathematical properties of ratio statistics to systematically inflate gene-level{pi} N /{pi}S estimates even in the absence of selection. Notably, filtering this rare variation reveals previously masked candidates for selection, including well-characterized antigens such as merozoite surface proteins. This framework provides a foundation for interpreting genomic data in pathogens with high variance in reproductive success. Significance StatementPathogen genomics increasingly guides public health decisions, yet organisms with complex life cycles can produce genealogies that violate assumptions of standard population genetic methods. Here I show that the malaria parasite Plasmodium falciparum has an excess of rare variation, perhaps produced from overdispersed distributions of reproductive success. These genealogies generate systematic biases in commonly reported statistics arising from the interaction between rare variation and the mathematical properties of the statistics themselves, not from biological processes often invoked to explain anomalous patterns. Correcting for this bias reveals signatures of selection in well-characterized vaccine candidates that were previously masked, demonstrating that appropriate null models are essential for identifying targets of intervention and inferring evolutionary history in pathogens and other systems with reproductive skew.