Here we describe a novel method, association of covariance for detecting differential co-expression (ACDC), to detect differential co-expression across multiple binary, ordinal, or continuous phenotypes or exposures. While many methods exists for binary conditions and a few for greater than two, we are unaware of any module-based differential co-expression approaches designed to detect differences across continuous conditions or multiple types of conditions simultaneously. If these distributions are statistically significantly different from random in one condition and not the other, the module is considered differentially co-expressed. Then, pairwise correlation coefficients are used to created a distribution of co-expression for each module under two conditions. By cutting the trees determined by average-linkage hierarchical clustering at a user-defined threshold, genes are split into modules. One highly-cited method, CoXpress, determines differentially co-expressed modules given microarray data ( 3). Methods may also detect differential co-expression for gene pairs across the phenotype of interest and then apply post-hoc clustering methods to identify co-expressed modules. Generally, these module-based methods can be distinguished from one another by, (i) whether modules are defined by the user or the method, (ii) if differential co-expression is detected within or between modules, and (iii) how many conditions are assessed. These approaches can have good statistical power due to a reduction in “noise” ( 2), or unrelated variation of individual genes by collapsing related genes into a single feature.
Module-based differential co-expression methods incorporate information about gene connectivity, and assume that the genes within a module are correlated in the general population. Differential co-expression methods test for differences in gene covariances, and thus, such approaches may illuminate regulatory mechanisms not identified by differential expression analysis alone ( 1). However, differential expression methods study each gene independent of any other and therefore may not capture transcriptional differences due to changes in gene-gene relationships. Results suggest that both expression levels and covariances of ADORA3, ALOX15, and IDO1 are associated with asthma control.Ĭonclusion: ACDC is a flexible extension to existing methodology that can detect differential co-expression across varying external variables.ĭifferential expression analysis has long been used to test for differences in transcriptional dependencies across conditions, and may explain phenotypic variation in a population. Results: We report an application to two cohorts of asthmatic patients with varying levels of asthma control to identify associations between gene co-expression and asthma control test scores. Methods: Here, we propose a novel approach for detection of differential co-expression that simultaneously accommodates multiple phenotypes or exposures with binary, ordinal, or continuous data types. Current methods only allow for assessment of co-expression variation across a singular, binary or categorical exposure or phenotype, limiting the information that can be obtained from these analyses. 4Division of Pulmonary and Critical Care Medicine, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United Statesīackground: Existing module-based differential co-expression methods identify differences in gene-gene relationships across phenotype or exposure structures by testing for consistent changes in transcription abundance.3Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United States.2Division of Pulmonary Medicine, Boston Children's Hospital and Harvard Medical School, Boston, MA, United States.1Department of Population and Public Health Sciences, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States.Katelyn Queen 1, My-Nhi Nguyen 1, Frank D.
0 kommentar(er)
