Genome-wide search algorithms for identifying dynamic gene co-expression via Bayesian variable selection.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput gene expression data offers opportunities to study dynamic gene-gene interactions.
• Existing statistical methods face computational challenges due to the vast number of gene combinations.
• Dynamic gene interactions are crucial for biological system regulation and response to stimuli.

Purpose of the Study:

• To develop computationally efficient Bayesian variable selection approaches for identifying dynamic gene-gene interactions.
• To reduce the computational intensity in analyzing large genomic datasets.
• To identify significant dynamic gene co-expression changes using Bayesian multiple hypothesis testing.

Main Methods:

• Utilizing Bayesian variable selection with spike-and-slab priors to focus on promising gene combinations.
• Implementing a Bayesian multiple hypothesis testing procedure for robust detection of co-expression changes.
• Comparing proposed algorithms with existing exhaustive search heuristics via simulation studies.

Main Results:

• The proposed Bayesian approaches significantly reduce computational intensity compared to exhaustive methods.
• The algorithms effectively identify subsets of gene combinations exhibiting dynamic co-expression.
• Demonstrated application to The Cancer Genome Atlas (TCGA) breast cancer dataset.

Conclusions:

• Bayesian variable selection and multiple hypothesis testing offer efficient solutions for analyzing dynamic gene-gene interactions.
• These methods facilitate the exploration of gene co-expression patterns linked to clinical outcomes like overall survival.
• The approach is valuable for large-scale genomic data analysis in cancer research.