Quantitative gene set analysis generalized for repeated measures, confounder adjustment, and continuous covariates.

Area of Science:

• Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• Gene set analysis (GSA) is powerful for weak biological signals in gene expression data.
• Permutation tests, common in GSA, are unsuitable for complex designs like longitudinal studies.
• Linear mixed models (LMMs) can analyze complex data structures and adjust for variability.

Purpose of the Study:

• To generalize the QuSAGE GSA algorithm to incorporate LMMs, creating Q-Gen.
• To address the limitations of existing GSA methods in handling complex study designs.

Main Methods:

• Generalized the QuSAGE algorithm to develop Q-Gen.
• Incorporated estimation adjustments for LMMs within the GSA framework.
• Assessed Q-Gen's performance against QuSAGE in longitudinal and confounder-adjusted analyses.

Main Results:

• Original QuSAGE failed to control type-I error in complex designs.
• Q-Gen demonstrated statistical appropriateness for longitudinal and confounder-adjusted analyses.
• Q-Gen showed increased sensitivity in analyzing a longitudinal influenza study.

Conclusions:

• Q-Gen extends QuSAGE, enabling appropriate LMM application in GSA.
• Provides enhanced flexibility for statistical modeling of complex microarray data.
• Offers greater sensitivity for GSA in complex biological studies.