henhaobio
Sign in

BioinformaticsGeneticsPeer-reviewed journal article

MOIRE: A software package for the estimation of allele frequenciesand effective multiplicity of infection from polyallelic data

Maxwell MurphyiD, Bryan GreenhouseiD

PDFDOIPreprintCode
In-depth full-text digest available · Jump to digest ↓

Abstract

MotivationMalaria parasite genetic data can provide insight into parasite phenotypes, evolution, and transmission. However, estimating key parameters such as allele frequencies, multiplicity of infection (MOI), and within-host relatedness from genetic data is challenging, particularly in the presence of multiple related coinfecting strains. Existing methods often rely on single nucleotide polymorphism (SNP) data and do not account for within-host relatedness. ResultsWe present MOIRE (Multiplicity Of Infection and allele frequency REcovery), a Bayesian approach to estimate allele frequencies, MOI, and within-host relatedness from genetic data subject to experimental error. MOIRE accommodates both polyallelic and SNP data, making it applicable to diverse genotyping panels. We also introduce a novel metric, the effective MOI (eMOI), which integrates MOI and within-host relatedness, providing a robust and interpretable measure of genetic diversity. Extensive simulations and real-world data from a malaria study in Namibia demonstrate the superior performance of MOIRE over naive estimation methods, accurately estimating MOI up to 7 with moderate sized panels of diverse loci (e.g. microhaplotypes). MOIRE also revealed substantial heterogeneity in population mean MOI and mean relatedness across health districts in Namibia, suggesting detectable differences in transmission dynamics. Notably, eMOI emerges as a portable metric of within-host diversity, facilitating meaningful comparisons across settings when allele frequencies or genotyping panels differ. Compared to existing software, MOIRE enables more comprehensive insights into within-host diversity and population structure. AvailabilityMOIRE is available as an R package at https://eppicenter.github.io/moire/. Contactmm@maxmurphy.dev Supplementary informationSupplementary data are available at Bioinformatics online.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (2 Aug 2024) · CC BY 4.0

MOIRE, a Bayesian R package, jointly infers allele frequencies, multiplicity of infection (MOI) and within-host relatedness from SNP or polyallelic malaria genotypes, yielding a panel-portable effective MOI.

Naive analysis of polyclonal malaria infections biases allele frequencies (inflating heterozygosity) and underestimates multiplicity of infection, while prior tools (THE REAL McCOIL, coiaf, SNP-Slice) accept only SNPs and ignore within-host relatedness. MOIRE handles polyallelic markers such as microhaplotypes and microsatellites, models genotyping error, and explicitly estimates relatedness.

Key results

  1. ShownIn simulations, how well MOI was recovered depended strongly on marker diversity. SNP panels resolved MOI only up to about 4 strains, diverse polyallelic panels extended accurate recovery to about 7, and naive allele counting underestimated MOI throughout.

    Each simulated collection had 100 samples, with individual MOI drawn from zero-truncated Poisson rates 1, 3 and 5 (mean MOI 1.58, 3.16, 5.03). Error rates were 0.01 (false positive) and 0.1 (false negative). The naive estimator (second-highest allele count) was biased low, and with SNPs it can resolve at most 2 strains. · Estimation of multiplicity of infection…; Tables

  2. ShownMOIRE's allele frequency estimates were unbiased across synthetic and real-world panels, so its heterozygosity estimates were also unbiased. Naive estimation overstated rare alleles and understated common ones, which inflated heterozygosity.

    Synthetic panels: 100 SNPs, or 30 loci with 5, 10 or 20 alleles. Real-panel simulations: 24 SNP, 101 SNP and microhaplotype panels of 128, 165 and 233 loci across 12 MalariaGEN Pf7 populations. In fully synthetic runs, allele frequencies stayed well estimated as error rates varied from 0 to 0.1. Runs without relatedness showed minimal added bias. · Estimation of multiplicity of infection…

  3. ShownWith polyallelic panels, effective MOI (eMOI, which combines MOI with within-host relatedness) was recovered accurately even when MOI or relatedness alone was poorly identified. SNP panels resolved individual and population relatedness poorly and were sensitive to the prior.

    With error rates of 0.01 and 0.1, polyallelic panels recovered eMOI accurately in all conditions. SNP panels did relatively well up to eMOI 4, and their population-level eMOI was biased only at higher mean MOI. False positives sometimes led monoclonal samples with inferred MOI above 1 to be called very highly related. Coverage of 95% credible intervals for population mean MOI and relatedness was poor. · Effective MOI; Estimation of multiplicity of infection…; Population inference; Tables

  4. ShownPopulation mean eMOI stayed stable across parasite genetic backgrounds and genotyping panels. Naive mean MOI and the within-host fixation index (FWS) shifted with heterozygosity and panel, which limits comparisons between settings.

    400 infections across 4 conditions (zero-truncated Poisson MOI with lambda 1 or 3; low or high relatedness) were genotyped in silico under 12 regional Pf7 backgrounds and 5 panels, with error rates 0.01 and 0.1. Microhaplotype panels gave accurate mean eMOI. SNP panels were biased downward but consistently, suggesting they may still order eMOI correctly. · Metric stability across genetic backgrounds; Figure legends

  5. SuggestedReanalysis of a Namibian microsatellite study kept the district ranking of mean MOI. MOIRE gave higher mean MOI and lower heterozygosity than the naive estimates, and found high within-host relatedness. Facility-level runs showed heterogeneity, with Rundu district hospital the most diverse.

    2585 samples, 29 facilities, 4 districts, 26 microsatellite loci. Naive estimates of mean MOI were lower than MOIRE's across districts; Andara and Rundu ranked highest and Zambezi lowest. Relatedness IQR .61-.91; heterozygosity IQR .55-.85. Rundu district hospital: eMOI 4.3 (95% CI 4.18-4.4), MOI 7 (6.5-7.5), relatedness .47 (.43-0.51). · Application to a study in Northern Namibia; Figure legends

Methods, data and reproducibility

A Bayesian model with a latent true-genotype layer that separates sample-specific false positive and false negative calls from true allele presence. It jointly infers per-sample MOI, within-host relatedness (average genome fraction identical by descent) and per-locus allele frequencies, assuming unlinked loci. It is fit by Metropolis-Hastings MCMC with parallel tempering, written in Rcpp.

  • Fully synthetic simulations: panels of 100 SNPs, or 30 loci with 5, 10 or 20 alleles, under varied MOI, relatedness and error rates · n = 100 samples per simulated collection
  • Simulations under 5 real genotyping panels (24 SNP, 101 SNP; microhaplotype panels of 128, 165, 233 loci) parameterized by MalariaGEN Pf7 regional allele frequencies · Plasmodium (MalariaGEN Pf7 regional populations) · n = 12 regional populations; 100 samples per simulation
  • Metric-stability simulations: infections with fixed MOI and relatedness genotyped in silico across Pf7 backgrounds and panels · Plasmodium (MalariaGEN Pf7 regional populations) · n = 400 infections across 4 conditions
  • Previously collected malaria samples from northeastern Namibia genotyped at 26 microsatellite loci, analysed by district and by health facility · Plasmodium (species not stated in text) · n = 2585 samples; 29 health facilities; 4 health districts
Methods & tools
MOIRE (R package), R, Rcpp, Metropolis-Hastings MCMC with parallel tempering, MalariaGEN Pf7, Within-host infection fixation index (FWS), Naive allele frequency and MOI estimators, THE REAL McCOIL, coiaf, SNP-Slice (SNP-only prior methods; discussed, not benchmarked)
Software & environment
R package with Rcpp-implemented samplers. Analyses used default priors with 40 parallel tempered chains, 5000 burn-in steps and 10,000 samples thinned to 1000. No package or dependency versions are given.
Reproducibility gaps
No repository or accession is given for the Namibian genotype data or the simulation scripts. Likelihood, prior, sampler and simulation details are only in the supplementary material.
Limitations (6)
  • The observation model uses only allele presence or absence. It ignores sequencing read counts, which could inform per-allele false positive rates and per-locus false negative rates. · stated by the authors
  • Each run assumes one well-mixed background population. Mixtures of parasite subpopulations and spatial structure are not modelled, so running Namibian samples by district assumes they share allele frequencies. · stated by the authors
  • Loci are assumed independent, with no linkage disequilibrium. The authors expect allele frequency and per-sample eMOI estimates to remain consistent under linkage provided loci are not systematically biased toward regions of high or low within-host relatedness. · stated by the authors
  • Credible interval coverage for population mean MOI and relatedness was poor. SNP panels could barely resolve relatedness and stayed sensitive to the uniform prior. · stated by the authors
  • False-positive calls can make monoclonal infections look like highly related polyclonal ones. Bias in individual MOI, relatedness and eMOI rose when false positive rates were higher and panel diversity was low. · stated by the authors
  • There is no head-to-head benchmark against THE REAL McCOIL, coiaf or SNP-Slice on shared SNP data; comparisons are only to naive estimators and FWS. The real-panel simulations also fixed error rates at 0.01 and 0.1. · noted in this summary

Read this if you estimate MOI, allele frequencies or within-host relatedness from malaria amplicon, microhaplotype or microsatellite panels and need a diversity metric you can compare across panels and settings.

Citations
35
FWCI
14.51
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISObama '26Obama '25Labbe '22Mehra '25Hashemi '24Taylor '19
74% semantic matchGenetics
Henri Christian Junior Tsoungui Obama, Kristan Alexander Schneider·PLoS Computational Biology·Jul 17, 2026