ASSIGN: context-specific genomic profiling of multiple heterogeneous biological pathways.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene-expression signature biomarkers often fail validation due to heterogeneous sample contexts and technical biases across platforms.
• This limits the elucidation of disease mechanisms and the development of personalized therapeutics.

Purpose of the Study:

• To present a novel pathway profiling toolkit, Adaptive Signature Selection and InteGratioN (ASSIGN), for robust and context-specific pathway analyses.
• To enable efficient capture of pathway activity in heterogeneous samples and across different profiling technologies.

Main Methods:

• Developed the ASSIGN framework utilizing a flexible Bayesian factor analysis approach.
• Enabled simultaneous profiling of multiple correlated pathways and adaptation of signatures to specific diseases.

Main Results:

• Demonstrated the robustness and versatility of ASSIGN in estimating pathway activity using simulated data, cell lines, and primary tissue samples.
• Validated ASSIGN's performance on The Cancer Genome Atlas breast carcinoma samples and genotoxic carcinogen-exposed liver samples.

Conclusions:

• ASSIGN provides a robust method for pathway analysis, overcoming challenges in biomarker validation.
• The toolkit facilitates better understanding of disease mechanisms and supports personalized therapeutic strategies.