Identifying genetic variants that might have significant joint effects on multifactorial disease phenotypes is now in great interest as single locus approach failed due to weak marginal effects. Malaria is a multifactorial disease occurring after infection by Plasmodium parasites. Frequency and severity of illness depends on known individual and environmental factors like age, sex and transmission intensity; but also on unknown genetic aspects. Thus, to determine the susceptibility or resistance of individuals to uncomplicated malaria, longitudinal surveys are useful as they allow finding confirmed individual tendencies based on several sampling. Here, we studied data from a long-term epidemiological and genetic survey of malaria disease in two family-based cohorts in Senegal.
The objective of this work was to take individual malaria tendencies (susceptibility or resistance) as phenotype and then test its association to a set of loci jointly.
We used the number Plasmodium falciparum malaria attacks (PFA) per trimester, a malaria trait known to be influenced by human genetics, to represent the frequency and severity of the outcome of infection. We then performed Generalized Linear Mixed Models that account for correlated random effects such as those due to genetic relationships among individuals, repeated measures and environmental covariates to extract individual tendencies.
We then performed genetic studies focusing on candidate genes for susceptibility/resistance to malaria. We used family-based methods with a multi-locus model, more powerful and better adapted for multifactorial diseases such as malaria, to test for genetic linkage and association at any number of independent loci simultaneously. Simulation studies showed a gain of power from single locus to multi-locus models in detecting genetic effects on a phenotype suspected to be influenced by several independent loci. Then, multi-locus models should be appropriate for malaria phenotypes supposed to be influenced by actions from many different genes having weak marginal effects. We used 45 Single Nucleotide Polymorphisms (SNPs) on candidate genes as genetic variables and the individual PFA tendencies as phenotype of interest. We then applied this method analyzing the SNPs one by one in a first step and SNPs showing at least a weak significance (P-value ≤ 0.10) for association with the phenotype were selected in a second step for a multi-locus model that analyzes simultaneous transmission of alleles from those SNPs. Five SNPs showed weak marginal protective effects against malaria after correction for multiple testing: three SNPs on the SLC4A1 gene located on chromosome 17 (ae1_20_21, P = 0.0005; ae1_117_118, P = 0.0598; ae1_174_187, P = 0.0995), one SNP on the γ-globin gene (Xmn1) located on chromosome 11 (Xmn1, P = 0.0598) and one other on the gene ABO located on chromosome 9 (abo297, P = 0.0854). We then analyzed these five loci together and obtained many significant joint effects (P-values distributed from 10-2 to 10-8 for joint effects corresponding to different ways of combining these five loci). This result shows how important our method is to identify sets of genes playing important roles in malaria and in multifactorial diseases in general.
