Uncovering Effective Explanations for Interactive Genomic Data Analysis.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Identifying discriminative features in multi-class genomic datasets is crucial for biological research.
• Existing tools struggle with the computational demands of large-scale genomic data analysis.

Purpose of the Study:

• To develop an interactive tool, GENVISAGE, for rapid identification and visualization of feature-based explanations in genomic datasets.
• To optimize the discovery of discriminative feature pairs for improved scalability and responsiveness.

Main Methods:

• Development of GENVISAGE, an interactive exploration tool for genomic data.
• Implementation of a suite of optimizations to accelerate the discovery of discriminative feature pairs.
• Benchmarking against competitive baselines to quantify performance improvements.

Main Results:

• GENVISAGE achieves a 400x speedup over baseline methods on multiple biological datasets.
• The tool successfully identifies literature-supported gene pairs discriminating chemotherapy drug treatments.
• Demonstrated scalability for large-scale genomic datasets.

Conclusions:

• GENVISAGE provides a rapid and interpretable solution for feature-based explanation generation in genomics.
• The tool's optimizations enhance responsiveness for large datasets, enabling real-time analysis.
• GENVISAGE has broad applicability to massive sequencing efforts and other scientific domains.