Across a marked gradient in transmission intensity among West African locations sampled here, the parasite populations show remarkably similar population genetic structure. There was a high allelic diversity for the ten P. falciparum microsatellite loci in all the eight locations, the overall H
index for each site ranging from 0.72 to 0.80, the same range as that previously seen in most other endemic African countries [8, 16, 33]. Although slightly lower diversity has been reported in a low-endemic area of Dakar in Senegal  and a highland area of western Kenya , a much lower diversity has only been reported for one African location (H
of 0.41 in Djibouti which has unstable epidemic transmission) .
As expected, the mean numbers of P. falciparum genotypes detected per infection varied among the populations studied, with highly mixed infections at each of the sites in Guinea which experience high transmission for much of the year, in contrast with sites having limited seasonal transmission further north in the region. This substantial difference is likely to be due to differences in endemicity as predicted, rather than confounding due to differences in sampling. Samples from five of the sites were from clinical cases presenting to health facilities, and those from the remaining three sites were from community sampling of asymptomatic individuals, but the latter sites were in the middle of the range geographically and in terms of proportions of mixed genotype infections, so this did not cause the observed north–south contrast in values. Review of comparisons elsewhere between asymptomatic and symptomatic infections also shows no consistent overall difference in proportions of mixed genotype infections [35–40].
The multi-locus index of association analysis initially indicated non-random patterns in several populations, but this was shown to be entirely due to only one or a few pairs of virtually identical isolates. In South American and Southeast Asian populations of low endemicity, identical P. falciparum isolates are sometimes seen as multi-locus genotypes persist through a number of self-fertilisation and transmission cycles. This has allowed comparisons of identical versus non-identical pairs of clones for estimation of phenotype heritability in a manner analogous to a conventional twin-pair study [41, 42]. In Africa, related parasites may occur within isolates but it is rare to see identical genotypes in different infections . It has been previously shown in The Gambia that identical parasite genotypes were more common in pairs of children sleeping in the same house who presented with malaria on the same day , and occasionally in children living close to each other within small villages . Such identical genotypes probably result from single mosquitoes infecting more than one individual or from transmission of parasites by different mosquitoes that fed on a single genotype gametocyte carrier.
As was seen here, it has elsewhere been shown that inclusion of a very small number of closely related parasite genotypes can generate a significantly non-random multi-locus index of association, even when there is no linkage disequilibrium among the loci in the population generally . Such occasional occurrence of identical parasites should not lead to a population being considered ‘clonal’, as such a term is not generally applied to human population genetic structure despite the presence of monozygotic twins. Instead, it has been suggested that the presence of many pairs of genotypically similar isolates may be taken as evidence of an ‘epidemic’ population structure [8, 46], but here there were very few such pairs and none of the populations could be described as having such a structure. It should be noted that if multiple sibling parasite genotypes within the same host were included in a crude analysis, it could give false appearance of linkage disequilibrium [47, 48]. Therefore, such a problem is avoided by restricting analysis to one parasite genotype per host [49–51], following the general principle whereby genotypes from closely related family members are not separately counted in studies of human linkage disequilibrium.
Over the sampled range of up to ~ 1200 kilometers between sites in West Africa, this microsatellite analysis found very low levels of genetic differentiation between the local populations of P. falciparum, with most pairwise F
ST values being less than 0.03. This is considerably less than the differentiation among local sites within non-African countries that are less endemic, including Malaysia , Papua New Guinea , the Philippines , and Brazil , with F
ST values exceeding 0.10 between sites separated by similar distances. It is instead consistent with previous sampling from more widely separated African populations, which has shown F
ST values of less than 0.05 for a similar set of microsatellite loci [8, 16]. Despite the ecological and epidemiological diversity in West Africa, there is likely to be considerable mixing of parasites between different locations, due to frequent movement of humans in this region . In comparison with South America and Southeast Asia, it will be difficult to identify discrete endemic locations in West Africa where malaria elimination might be achieved and sustained in the face of local migration.
Despite studying only ten polymorphic loci and sampling a very limited number of isolates from some of the sites, marked variation in proportions of multiple clone infections and relatively similar genetic structure of P. falciparum populations has been clearly shown, providing a framework for future genomic-scale studies. Comprehensive analysis of genome sequence variation should allow finer differences in population structure to be detected , including variation in patterns across the genome and identification of genes under natural selection, particularly with large sample sizes. Given the high levels of recombination and minimal reproductive isolation of parasite populations in West Africa, we predict that differential signatures of selection in particular populations will be detectable against a background of neutral genomic variation that is more spatially homogeneous.