GPAR

GPAR predicts drug mechanisms of action (MOAs) from drug-induced gene expression profiles using deep learning to enable mechanistic inference and hypothesis generation.

Key Features:

• Deep learning integration: Applies deep learning and machine learning algorithms to model relationships between gene expression changes and drug MOAs.
• Drug-induced gene expression profiling: Uses large-scale drug-induced gene expression profiles as input for MOA modeling and prediction.
• Customizable training sets: Trains models on user-defined/customized datasets to focus on specific biological questions or drug classes.
• Performance evaluation and prediction: Evaluates model performance with cross-validation and reports improved MOA prediction relative to Gene Set Enrichment Analysis (GSEA).

Scientific Applications:

• Drug repurposing: Predicts MOAs to identify potential new therapeutic applications for existing compounds.
• Hypothesis generation for experimental validation: Generates mechanistic hypotheses from gene expression–based MOA predictions to guide follow-up experiments.
• High-throughput gene expression analysis: Analyzes large-scale/high-throughput gene expression datasets for systematic investigation of drug-induced biological effects.

Methodology:

Queries drug-induced gene expression profiles and applies deep learning/machine learning algorithms to train models on customized datasets, with model evaluation performed via cross-validation.