Markov chain Monte Carlo Gibbs sampler approach for estimating haplotype frequencies among multiple malaria infected human blood samples

Table 1 The correlation (R²) of the estimated haplotype frequencies with simulated population and sample haplotype frequencies across statistical methods and four conditions of LoD_SNP (30%, 20%, 10%, 0%) and LoD_MOI (15%, 10%, 5%, 0%)

	MHF	R-EM	Bayesian	EM	MCMC	Gibbs
Population haplotype (LoD_SNP/LoD_MOI)
0/0	0.949	0.913	0.955	0.977	0.973	0.961
0.10/0.05	0.953	0.932	0.960	0.978	0.980	0.970
0.20/0.10	0.962	0.949	0.960	0.977	0.981	0.975
0.30/0.15	0.957	0.962	0.960	0.974	0.976	0.977
Sample haplotype (LoD_SNP/LoD_MOI)
0/0	0.960	0.926	0.962	0.983	0.978	0.966
0.10/0.05	0.963	0.944	0.968	0.985	0.986	0.976
0.20/0.10	0.971	0.960	0.968	0.984	0.987	0.982
0.30/0.15	0.967	0.972	0.967	0.982	0.983	0.984

Higher value represents higher accuracy
MHF MalHaploFreq, R-EM malaria em, Bayesian Bayesian statistic, EM EM algorithm, MCMC Markov chain Monte Carlo, Gibbs Gibbs sampler, LoD limit of detection, SNP single nucleotide polymorphisms, MOI multiplicity of infection

ISSN: 1475-2875