Testing differentially expressed genes in dose-response studies and with ordinal phenotypes.

Area of Science:

• Genomics
• Statistical Bioinformatics
• Computational Biology

Background:

• Testing for differentially expressed genes (DEGs) is crucial in biological research, particularly with multiple experimental groups.
• Existing methods often require assumptions about the relationship between group levels (e.g., dose-response, disease stages), limiting their applicability.
• Small sample sizes or a large number of groups can pose challenges for traditional DEG analysis.

Purpose of the Study:

• To introduce and evaluate a novel statistical test for identifying DEGs between more than two ordered groups.
• To leverage the ordering of groups (e.g., dose levels, disease stages) without imposing restrictive assumptions like monotonicity.
• To provide a robust method applicable to scenarios with many dose levels or few subjects per group.

Main Methods:

• The proposed method utilizes a mixed effects model framework.
• It tests for zero variance components to detect differential expression across ordered levels.
• The approach borrows strength across groups, enhancing power in challenging sample sizes or group numbers.

Main Results:

• The new test demonstrated high speed and power in simulation studies.
• Performance was validated on several publicly available biological datasets.
• Comparisons showed favorable results against alternative testing procedures.

Conclusions:

• The developed mixed model-based test offers a fast, powerful, and flexible approach for DEG analysis in ordered groups.
• It effectively handles scenarios with many levels or small sample sizes without strict assumptions on the dose-response or phenotype relationship.
• The implementation is publicly available in R, facilitating broader adoption in the scientific community.