Non-linear tests for identifying differentially expressed genes or genetic networks

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• The Hotelling T2 statistic is used for identifying differentially expressed genetic networks.
• Its power is limited due to its quadratic form of gene expression differences.

Purpose of the Study:

• To develop a general statistical framework for constructing non-linear tests.
• To improve the power of statistical tests for differential gene expression analysis.

Main Methods:

• Developed two novel non-linear test statistics using non-linear transformations of gene expression means.
• Derived asymptotical distributions under null and alternative hypotheses.
• Applied new statistics to real biological datasets.

Main Results:

• Non-linear test statistics demonstrated higher power than the Hotelling T2 statistic under certain conditions.
• P-values from non-linear statistics were significantly smaller than those from T2.
• Simulations confirmed Type I errors align with thresholds and statistics fit chi2 distribution.

Conclusions:

• The developed non-linear test statistics offer a powerful alternative for detecting differential gene expression.
• These methods enhance the analysis of genetic networks compared to existing approaches.