Association of rare haplotypes on ULK4 and MAP4 genes with hypertension

Abstract

Several variants have been implicated earlier on ULK4 and MAP4 genes on chromosome 3 to be associated with hypertension. As a natural follow-up step, we explore association of haplotypes in those genes. We consider the Genetic Analysis Workshop 19 real data on unrelated individuals and analyze haplotype blocks of 5 single-nucleotide polymorphisms through a sliding window approach. We apply 4 haplotype association methods—haplo.score, haplo.glm, hapassoc, and logistic Bayesian LASSO (LBL)—and for comparison, sequence kernel association test (SKAT) and its variants. We find several rare haplotype blocks to be associated. To get an idea about the false-positive proportions, we also analyzed the data after permuting the case-control status of individuals. We found that LBL, unlike the other methods, maintains low false-positive rates in presence of rare haplotypes. Thus, we conclude that the haplotypes found to be associated by LBL are more likely to be true positive. SKAT and its variants did not find significance on either gene.

Comparison of haplotype-based statistical tests for disease association with rare and common variants

“…From a biological point of view, haplotype effects can be expected in many real data sets, and in such cases, our work shows that just collapsing SNPs together does not retain same power as analyzing them as a haplotype block. Similar results have been recently shown elsewhere [11,12]; however, we add a disclaimer that comparison of haplotyping versus collapsing approaches is not the focus here and thus has not been investigated in a more rigorous manner. We would like to refer readers interested in this comparison to [9,10,11,18] wherein some other collapsing approaches are also considered.…”

“…On the other hand, BhGLM is highly conservative with almost all type I error rates close to 0. LBL, with the cutoff of 2, has controlled type I error rates, each below 5%, consistent with previous studies [12,14,15,26]. Similar trend is seen for other sample sizes (see Supplementary Tables S1 and S2).…”

“…In addition to LBL, several other haplotype-based methods were able to uncover association, illustrating aptly that haplotype-based methods work well with not only GWAS data but also sequence data, contrary to popular belief. This was also demonstrated in GAW19 sequence data analysis recently [12]. Moreover, individual haplotype-specific methods provide additional information about the specific haplotypes that are associated and the direction of their effects.…”

“…To get an idea about whether the association signals found by various methods are likely to be true or false signals, we also analyzed the same windows by permuting the affection status and averaging the number of significant results over all 109 windows (as in [12]). The type I error rates were estimated to be 1.35%, 34.98%, 3.59%, 3.67% and 0.92% for LBL, Haplo.glm, Haplo.score individual haplotype-specific and global tests and rGLM, respectively, while for the rest it was 0.…”

“…Haplotype association has a rich literature with many earlier studies highlighting their importance including their increased power than single-SNP approaches when there are multiple causal variants acting in cis or if the causal variant(s) are not genotyped [6][7][8]. More recent works have shown that rHTVs formed by common SNPs are more likely to tag rSNVs and that rHTV association methods may be more powerful than popular collapsing methods for detecting rSNVs in some situations [9][10][11][12].…”