Modeling genetic polymorhphisms and sickle cell associated vasoocclusive events using classification and regression trees (CART)

Purpose Although sickle cell anemia is a single gene (HBB) disorder, the variable pattern of complications seen among patients is unlikely the result of solely the HBB mutation. Conventional case-control analyses using single nucleotide polymorphisms (SNPs) are useful in looking for single gene associations, however studying gene interactions becomes increasingly inefficient as the number of SNPs increase. Application of more complex statistical methods such as classification and regression trees allows for the analysis of many SNPs and covariates simultaneously. Methods From the over 4000 patients in the Cooperative Study for Sickle Cell Disease, a subset of 1584 were genotyped for 180 SNPs in 92 genes. Clinical data for these patients were merged with genotype data and 576 patients were identified who had at least one of the following vasoocclusive events: stroke, osteonecrosis of the humeral or femoral head and/or priapism. With the remaining patients serving as controls, CART was run to identify genes and covariates whose interactions characterize patients with vasoocclusive events. Results The final tree, after pruning, shows five terminal nodes that predict vasoocclusive event. Along with age, the genes ANXA2, BMP6, SELP, TGFBR2 and TGFBR3 were found to be associated with the vasoocclusive event phenotype; each of these genes have been found to be associated with at least one of the individual phenotypes of stroke, osteonecrosis and/or priapism previously. Conclusion These results support our parallel work using Bayesian networks to understand the interactions among genes underlying the clinical heterogeneity of sickle cell anemia and studies of individual phenotypes. While Bayesian networks have superior predictive power, CART provides a simple initial screening of genes and covariates that may be simultaneously associated with the phenotype.