BADGER: biologically-aware interpretable differential gene expression ranking model.

Area of Science:

• Computational biology
• Genomics
• Pharmacology

Background:

• Identifying genes affected by drugs is vital for drug repurposing.
• Analyzing gene expression changes reveals drug-impacted genes.

Purpose of the Study:

• To develop an interpretable model for predicting gene expression changes from drug-cell interactions.
• To enhance drug repurposing by understanding drug effects on cancer cell lines.

Main Methods:

• The Biologically-Aware Interpretable Differential Gene Expression Ranking (BADGER) model integrates prior knowledge of drug targets via pathway information.
• It handles novel cancer cell lines using similarity-based embedding.
• Three attention blocks mimic cascading chemical compound effects.

Main Results:

• The BADGER model demonstrates superior performance over baselines in predicting gene expression changes.
• It shows robust prediction capabilities for untested drug-cell line combinations.
• The model overcomes limitations in cell line range and explainability.

Conclusions:

• The BADGER model is valuable for drug repurposing, especially for new or resistant diseases.
• It facilitates developing therapeutic plans by leveraging disease similarities.
• The model enhances understanding of drug-cell responses.