Adaptively Integrative Association between Multivariate Phenotypes and Transcriptomic Data for Complex Diseases

Phenotype–gene association studies can uncover disease mechanisms for translational research. Association with multiple phenotypes or clinical variables in complex diseases has the advantage of increasing statistical power and offering a holistic view. Existing multi-variate association methods mostly focus on SNP-based genetic associations. In this paper, we extend and evaluate two adaptive Fisher’s methods, namely AFp and AFz, from the p-value combination perspective for phenotype–mRNA association analysis. The proposed method effectively aggregates heterogeneous phenotype–gene effects, allows association with different data types of phenotypes, and performs the selection of the associated phenotypes. Variability indices of the phenotype–gene effect selection are calculated by bootstrap analysis, and the resulting co-membership matrix identifies gene modules clustered by phenotype–gene effect. Extensive simulations demonstrate the superior performance of AFp compared to existing methods in terms of type I error control, statistical power and biological interpretation. Finally, the method is separately applied to three sets of transcriptomic and clinical datasets from lung disease, breast cancer, and brain aging and generates intriguing biological findings.

Recent advances and challenges of rare variant association analysis in the biobank sequencing era

Causal variants for rare genetic diseases are often rare in the general population. Rare variants may also contribute to common complex traits and can have much larger per-allele effect sizes than common variants, although power to detect these associations can be limited. Sequencing costs have steadily declined with technological advancements, making it feasible to adopt whole-exome and whole-genome profiling for large biobank-scale sample sizes. These large amounts of sequencing data provide both opportunities and challenges for rare-variant association analysis. Herein, we review the basic concepts of rare-variant analysis methods, the current state-of-the-art methods in utilizing variant annotations or external controls to improve the statistical power, and particular challenges facing rare variant analysis such as accounting for population structure, extremely unbalanced case-control design. We also review recent advances and challenges in rare variant analysis for familial sequencing data and for more complex phenotypes such as survival data. Finally, we discuss other potential directions for further methodology investigation.

A Comparison of Methods for Gene-Based Testing That Account for Linkage Disequilibrium

Controlling the type I error rate while retaining sufficient power is a major concern in genome-wide association studies, which nowadays often examine more than a million single-nucleotide polymorphisms (SNPs) simultaneously. Methods such as the Bonferroni correction can lead to a considerable decrease in power due to the large number of tests conducted. Shifting the focus to higher functional structures (e.g., genes) can reduce the loss of power. This can be accomplished via the combination of p-values of SNPs that belong to the same structural unit to test their joint null hypothesis. However, standard methods for this purpose (e.g., Fisher’s method) do not account for the dependence among the tests due to linkage disequilibrium (LD). In this paper, we review various adjustments to methods for combining p-values that take LD information explicitly into consideration and evaluate their performance in a simulation study based on data from the HapMap project. The results illustrate the importance of incorporating LD information into the methods for controlling the type I error rate at the desired level. Furthermore, some methods are more successful in controlling the type I error rate than others. Among them, Brown’s method was the most robust technique with respect to the characteristics of the genes and outperformed the Bonferroni method in terms of power in many scenarios. Examining the genetic factors of a phenotype of interest at the gene-rather than SNP-level can provide researchers benefits in terms of the power of the study. While doing so, one should be careful to account for LD in SNPs belonging to the same gene, for which Brown’s method seems the most robust technique.

The generalized Fisher's combination and accurate p-value calculation under dependence

Combining dependent tests of significance has broad applications but the related p-value calculation is challenging. For Fisher's combination test, current p-value calculation methods (e.g., Brown's approximation) tend to inflate the type I error rate when the desired significance level is substantially less than 0.05. The problem could lead to significant false discoveries in big data analyses. This paper provides two main contributions. First, it presents a general family of Fisher type statistics, referred to as the GFisher, which covers many classic statistics, such as Fisher's combination, Good's statistic, Lancaster's statistic, weighted Z-score combination, etc. The GFisher allows a flexible weighting scheme, as well as an omnibus procedure that automatically adapts proper weights and the statistic-defining parameters to a given data. Second, the paper presents several new p-value calculation methods based on two novel ideas: moment-ratio matching and joint-distribution surrogating. Systematic simulations show that the new calculation methods are more accurate under multivariate Gaussian, and more robust under the generalized linear model and the multivariate t-distribution. The applications of the GFisher and the new p-value calculation methods are demonstrated by a gene-based SNP-set association study. Relevant computation has been implemented to an R package GFisher available on the Comprehensive R Archive Network. This article is protected by copyright. All rights reserved.

Meta-Analyzing Multiple Omics Data With Robust Variable Selection

High-throughput omics data are becoming more and more popular in various areas of science. Given that many publicly available datasets address the same questions, researchers have applied meta-analysis to synthesize multiple datasets to achieve more reliable results for model estimation and prediction. Due to the high dimensionality of omics data, it is also desirable to incorporate variable selection into meta-analysis. Existing meta-analyzing variable selection methods are often sensitive to the presence of outliers, and may lead to missed detections of relevant covariates, especially for lasso-type penalties. In this paper, we develop a robust variable selection algorithm for meta-analyzing high-dimensional datasets based on logistic regression. We first search an outlier-free subset from each dataset by borrowing information across the datasets with repeatedly use of the least trimmed squared estimates for the logistic model and together with a hierarchical bi-level variable selection technique. We then refine a reweighting step to further improve the efficiency after obtaining a reliable non-outlier subset. Simulation studies and real data analysis show that our new method can provide more reliable results than the existing meta-analysis methods in the presence of outliers.