Identification of differentially expressed gene sets using the Generalized Berk-Jones statistic

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Cancer genomics studies often analyze single genes for differential expression, leading to reproducibility challenges.
• Distinguishing causative genomic alterations from downstream effects in transcriptomic data is a significant analytical hurdle.
• There is a need for robust, well-powered analytical methods for reproducible identification of causative genes in cancer.

Purpose of the Study:

• To propose and evaluate a set-based statistical procedure for robust differential gene expression testing.
• To adapt the Generalized Berk-Jones (GBJ) statistic for identifying transcription factors in estrogen receptor-positive breast cancer.
• To enhance the reproducibility and consistency of findings across multiple transcriptomic datasets.

Main Methods:

• Developed a set-based statistical approach aggregating information from multiple correlated genomic markers.
• Adapted the Generalized Berk-Jones statistic for differential expression analysis of transcription factors.
• Validated the method's reproducibility by applying it to 21 diverse breast cancer datasets.

Main Results:

• The proposed set-based GBJ method demonstrated improved consistency compared to existing algorithms (Generalized Higher Criticism, Gene Set Analysis, Gene Set Enrichment Analysis).
• Consistent identification of significant transcription factors was achieved across the majority of the 21 analyzed datasets.
• The approach effectively aggregates information, leading to more robust and reproducible results in cancer genomics.

Conclusions:

• The set-based Generalized Berk-Jones approach offers a more robust and reproducible method for differential gene expression analysis in cancer genomics.
• This method enhances the ability to identify potentially causative genes and transcription factors, improving our understanding of cancer etiology.
• The findings highlight the utility of set-based analyses for overcoming limitations in transcriptomic study reproducibility.