Dcifer: an IBD-based method to calculate genetic distance between polyclonal infections

Abstract

An essential step toward reconstructing pathogen transmission and answering epidemiologically relevant questions from genomic data is obtaining pairwise genetic distance between infections. For recombining organisms such as malaria parasites, relatedness measures quantifying recent shared ancestry would provide a meaningful distance, suggesting methods based on identity by descent (IBD). While the concept of relatedness and consequently an IBD approach is fairly straightforward for individual parasites, the distance between polyclonal infections, which are prevalent in malaria, presents specific challenges and awaits a general solution that could be applied to infections of any clonality and accommodate multiallelic (e.g. microsatellite or microhaplotype) and biallelic (SNP) data. Filling this methodological gap, we present Dcifer (Distance for complex infections: fast estimation of relatedness), a method for calculating genetic distance between polyclonal infections, which is designed for unphased data, explicitly accounts for population allele frequencies and complexity of infection, and provides reliable inference. Dcifer’s IBD-based framework allows us to define model parameters that represent interhost relatedness and to propose corresponding estimators with attractive statistical properties. By using combinatorics to account for unobserved phased haplotypes, Dcifer is able to quickly process large datasets and estimate pairwise relatedness along with measures of uncertainty. We show that Dcifer delivers accurate and interpretable results and detects related infections with statistical power that is 2-4 times greater than that of approaches based on identity by state. Applications to real data indicate that relatedness structure aligns with geographic locations. Dcifer is implemented in a comprehensive publicly available software package.

Highly multiplexed ddPCR-amplicon sequencing reveals strongPlasmodium falciparumpopulation structure and isolated populations amenable to local elimination efforts in Zanzibar

“… A , Histogram of pairwise IBD. Pairwise IBD between all samples, estimated by Dcifer 42 . Inset shows the heavy tail of the distribution, with some pairs of samples having IBD ≥ 0.9.…”

“…B , Histogram of pairwise IBD. Pairwise IBD between all samples from Zanzibar (n=290), with mainland Tanzania (n=80), Kenya (n=35) and themselves, estimated by Dcifer 42 . Note that the y-axis is on square-root scale.…”

“…We estimated pairwise genetic relatedness between infections using Dcifer , an IBD-based method to infer the degree of shared ancestry between polyclonal infections 42 , implemented in R package dcifer , version 1.2.0 (https://cran.r-project.org/web/packages/dcifer/index.html). Dcifer allows for unphased multiallelic data such as microhaplotypes, explicitly accounts for MOI and population allele frequencies, and does not require densely distributed or linked markers.…”

“…We estimated pairwise genetic relatedness between infections using Dcifer, an IBD-based method to infer the degree of shared ancestry between polyclonal infections 42 , implemented in R package dcifer, version 1.2.0 (https://cran.r-project.org/web/packages/dcifer/index.html).…”

“…This provides a metric at finer spatiotemporal scales (10s to 100s of kilometers; weeks to months) for characterizing changes and measuring connectivity in parasite populations than classical population genetic measures of population diversity and divergence (e.g., π or F ST ) 14,21 , on timescales relevant to malaria elimination efforts 39 . While many IBD-based methods use whole genome sequencing (WGS) data and are limited to monoclonal infections or to biallelic loci 40,41 , a recently developed approach allows to infer the level of shared common ancestry (i.e., relatedness) between polyclonal infections from unphased multiallelic data such as microhaplotype data generated by AmpSeq 42 . Measures such as population genetic diversity, linkage disequilibrium, or multiplicity of infection (MOI) might be useful as surrogate markers of changes in transmission intensity 7,18,31 .…”