A PLSPM-Based Test Statistic for Detecting Gene-Gene Co-Association in Genome-Wide Association Study with Case-Control Design

Abstract

For genome-wide association data analysis, two genes in any pathway, two SNPs in the two linked gene regions respectively or in the two linked exons respectively within one gene are often correlated with each other. We therefore proposed the concept of gene-gene co-association, which refers to the effects not only due to the traditional interaction under nearly independent condition but the correlation between two genes. Furthermore, we constructed a novel statistic for detecting gene-gene co-association based on Partial Least Squares Path Modeling (PLSPM). Through simulation, the relationship between traditional interaction and co-association was highlighted under three different types of co-association. Both simulation and real data analysis demonstrated that the proposed PLSPM-based statistic has better performance than single SNP-based logistic model, PCA-based logistic model, and other gene-based methods.

Using composite phenotypes to reveal hidden physiological heterogeneity and model oxygen saturation variation of high altitude acclimatization in a Chinese Han longitudinal cohort

“…combinations of physiological phenotypes could become a promising alternative 13–15 . There are several methods to extract composite phenotypes from multiple traits, such as Principal Component Analysis (PCA)-based methods 14,16,17 and Partial Least Squares (PLS)-based methods 9,18,19 . PLS-based methods have better performance than PCA-based methods 18,19 .…”

“…There are several methods to extract composite phenotypes from multiple traits, such as Principal Component Analysis (PCA)-based methods 14,16,17 and Partial Least Squares (PLS)-based methods 9,18,19 . PLS-based methods have better performance than PCA-based methods 18,19 . Partial Least Squares Path Modeling (PLSPM) is the PLS-based approach to Structural Equation Modeling 20–22 , which can also be viewed as a method for analyzing multiple relationships between groups of variables.…”

“…Partial Least Squares Path Modeling (PLSPM) is the PLS-based approach to Structural Equation Modeling 20-22 , which can also be viewed as a method for analyzing multiple relationships between groups of variables. In the PLSPM framework, there are generally two ways to define composite phenotypes, i.e., latent variables 9,19-22 : one is using the prior knowledge and the other is using data-driven methods such as spectral clustering 23,24 .…”

A gene-based information gain method for detecting gene–gene interactions in case–control studies

“…There are some gene-based GGI detecting methods, such as canonical correlation-based U-statistic model, 21 sparse canonical correlation analysis model, 22 kernel canonical correlation-based U-statistic model (KCCU), 23,24 kernel regression model (KR), 25 partial least squares path model (PLSPM and mPLSPM) 26,27 and so on. However, most of these methods can only reflect the linear relationship between two genes, and cannot reflect the nonlinear relationship.…”

From Interaction to Co-Association —A Fisher r-To-z Transformation-Based Simple Statistic for Real World Genome-Wide Association Study

“…While traditional LRT only provide one way to identify the part under the nearly independent condition, with less power for the left attributed to the correlation. To solve this problem, in the context of a standard case-control design, three gene-based statistics, CCU [6], KCCU [7] and PLSPM-based statistic [17], have been developed in our former work based on the difference of correlation of two genes between cases and controls. Actually, similar idea has already been employed to develop new statistics recently [9], [11], [12].…”