Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix

Abstract

Correlated multiple testing is widely performed in genetic research, particularly in multilocus analyses of complex diseases. Failure to control appropriately for the effect of multiple testing will either result in a flood of false-positive claims or in true hits being overlooked. Cheverud proposed the idea of adjusting correlated tests as if they were independent, according to an 'effective number' (M eff ) of independent tests. However, our experience has indicated that Cheverud's estimate of the M eff is overly large and will lead to excessively conservative results. We propose a more accurate estimate of the M eff , and design M eff -based procedures to control the experiment-wise significant level and the false discovery rate. In an evaluation, based on both real and simulated data, the M eff -based procedures were able to control the error rate accurately and consequently resulted in a power increase, especially in multilocus analyses. The results confirm that the M eff is a useful concept in the error-rate control of correlated tests. With its efficiency and accuracy, the M eff method provides an alternative to computationally intensive methods such as the permutation test. Heredity (2005) 95, 221-227.

Genetic variation in the RANKL/RANK/OPG signaling pathway is associated with bone turnover and bone mineral density in men

“…The latter, to take account of multiple testing, was calculated by dividing 0.05 by the number of independent genotyped SNPs according to the method described by Li and Ji. (42) Correction for multiple testing…”

Multi-environment QTL mixed models for drought stress adaptation in wheat

“…In this paper, a Wald test was performed at 5-cM intervals along the genome; equivalent to 614 different tests for each trait. A modified Bonferroni correction was applied to take into account the correlations between tests (Li and Ji 2005). An alternative approach to reduce the number of multiple tests performed is to simultaneously fit all genetic predictors across the genome as random effects, with either a variance component per chromosome or a variance component for the whole genome, and use an outlier detection method to detect QTL in a nested iterative approach (Gilmour 2007;Verbyla et al 2007).…”

“…A Bonferroni-type correction adjusts the experiment-wide error rate for the number of tests performed but, since it assumes that all tests are independent (which they are not) this correction results in a too conservative threshold for QTL detection. The problem can be alleviated by approaches which determine the effective number of tests based on a principal component decomposition of the full set of explanatory variables (Cheverud 2001;Li and Ji 2005). In this study, the Li and Ji (2005) adjustment was calculated at a significance level of a = 0.10.…”

“…The problem can be alleviated by approaches which determine the effective number of tests based on a principal component decomposition of the full set of explanatory variables (Cheverud 2001;Li and Ji 2005). In this study, the Li and Ji (2005) adjustment was calculated at a significance level of a = 0.10. The Wald statistic P values were transformed to a -log 10 scale to produce QTL profiles for each linkage group.…”

A multiple testing correction method for genetic association studies using correlated single nucleotide polymorphisms

“…Cheverud [2001] was the first to propose this idea for multiple testing correction and published a formula for calculating M eff when SNP markers are correlated. However, Cheverud's M eff is still overly conservative when there is high LD among SNPs [Li and Ji, 2005;Salyakina et al, 2005]. Nyholt [2005] suggested excluding all SNPs in perfect LD except one prior to using Cheverud's M eff as a means to improve the adjustment accuracy, but this method remains overly conservative.…”