GSDA

GSDA detects complex non-monotonic associations between gene sets and categorical, quantitative, and censored event-time endpoints by generalizing distance correlations for gene-set association analysis.

Key Features:

• Generalized distance correlations: Extends distance correlation methodology to evaluate associations involving categorical, quantitative, and censored event-time variables.
• Non-monotonic relationship detection: Detects complex non-monotonic associations between gene sets and variables that can be missed by methods assuming monotonic relationships.
• Backward elimination for driver genes: Implements a backward elimination procedure to identify which genes within a set drive significant associations.
• Gene-set association focus: Evaluates empirical associations at the gene-set level to link pathways or gene sets with biological endpoints such as event-free survival.

Scientific Applications:

• Simulation benchmarking: Demonstrated superior performance in simulation studies compared to six published methods.
• Pediatric acute myeloid leukemia (AML) EFS analysis: Identified an association between event-free survival (EFS) and a 56-gene AML pathway gene set, narrowed the finding to five genes, and validated the association in an independent cohort.

Methodology:

Generalizes distance correlations to categorical, quantitative, and censored event-time endpoints and applies a backward elimination procedure to identify contributing genes; performance was evaluated via simulation studies and cohort validation.