Analysis of multiple SNPs in a candidate gene or region

Abstract

We consider the analysis of multiple SNPs within a gene or region. The simplest analysis of such data is based on a series of single SNP hypothesis tests, followed by correction for multiple testing, but it is intuitively plausible that a joint analysis of the SNPs will have higher power, particularly when the causal locus may not have been observed. However, standard tests, such as a likelihood ratio test based on an unrestricted alternative hypothesis, tend to have large numbers of degrees of freedom and hence low power. This has motivated a number of alternative test statistics. Here we compare several of the competing methods, including the multivariate score test (Hotelling's test) of Chapman et al. [2003], Fisher's method for combining p-values, the minimum p-value approach, a Fourier transform based approach recently suggested by Wang and Elston [2007] and a Bayesian score statistic proposed for microarray data by Goeman et al. [2005]. Some relationships between these methods are pointed out, and simulation results given to show that the minimum p-value and the Goeman et al. [2005] approaches work well over a range of scenarios. The Wang and Elston approach often performs poorly; we explain why, and show how its performance can be substantially improved.

Statistical tests for detecting associations with groups of genetic variants: generalization, evaluation, and implementation

“…We compare the performance of multiple-test statistics, including SUM: Sum test 11 HOI: Hotelling's T 2 -test assuming independence 20 STZ: Stouffer's Z-test 15 MDF: MultiDegree of Freedom test 16,17 These four tests are fully defined in Table 1. Note that HOI sets s to be the identity matrix, allowing for more direct comparisons with the other tests, which omit any reference to s. We focus on these four tests because they highlight the importance of the first two global features and because they are the most ubiquitous, noting that the recommended versions of tests using C-alpha, similarity regression, variance components, and kernels are all equivalent to MDF.…”

“…Standard rare-variant tests, such as the Sum test, 11 Hotelling's T 2 -test, 20 Stouffer's Z-test, 15 Data Adaptive Sum test, 21 C-alpha, 16 similarity regression, 17 variance components, 17 CMC, 22 and SKAT, 23 to name a few, only vary by their chosen set of weights. We specify five global features of the weights that vary among common test statistics.…”

“…9 Numerous statistical tests that can search for these regional associations have already been introduced, developed, and compared. 7,[10][11][12][13][14][15][16][17][18][19] Our three goals in this paper are to (1) Unify (2) Identify, and (3) Modify association tests for rare variants. First, we introduce a simple statistical framework and show that the majority of rare-variant association tests can be reformulated within this framework.…”

Permutation-based approaches do not adequately allow for linkage disequilibrium in gene-wide multi-locus association analysis

“…Chapman and Whittaker 18 also explored a Bayesian score test 20 as well as the method proposed by Wang and Elston. 21 However, the former is not currently implemented for GWAS, and the latter did not perform well, 18 so we did not investigate those approaches in the present study. Instead, we explored multi-marker logistic regression analysis.…”

“…Chapman and Whittaker 18 have previously reported a comparative study of the power of several statistical tests for combining SNPs in a candidate region. For comparison with the permutation-based tests, we studied the Hotelling's T 2 (H-T2) test, 19 which Chapman and Whittaker 18 also refer to as a 'multivariate score test' as it compares the differences between the multivariate means of two samples.…”