GENVISAGE

GENVISAGE identifies discriminative feature pairs in genomic datasets to distinguish biological classes (e.g., genes separating healthy versus diseased patients) and to reveal gene-pair interactions relevant to treatment response.

Key Features:

• Rapid discovery: Achieves ~400X improvement over competitive baselines to accelerate identification of top discriminative feature pairs in large datasets.
• Feature-pair focus: Evaluates pairs of features rather than individual features to capture interactions critical for distinguishing biological states.
• Scalability: Handles high-dimensional datasets with tens of thousands of objects (samples) and features (e.g., gene-expression measurements).
• Visualization: Produces visualizations corresponding to identified discriminative feature pairs to support interpretable analysis.

Scientific Applications:

• Drug response studies: Identifies pairs of genes whose transcriptomic responses discriminate between treatments with various chemotherapy drugs.
• Disease classification: Distinguishes genes that differentiate healthy from diseased patients in genomic analyses.

Methodology:

Employs a suite of computational optimizations and focuses on assessing discriminative feature pairs rather than individual features to enable analysis of large-scale genomic datasets, yielding approximately 400X speed improvement over competitive baselines.