DrDimont

DrDimont performs differential analysis of transcriptomics, proteomics, phosphosite, and metabolomics-derived molecular networks to predict and explain condition-specific drug responses.

Key Features:

• Multi-Omics Integration: Constructs condition-specific networks for each omics layer using molecular correlations, reduces them, and combines them into heterogeneous multi-omics networks.
• Novel Semi-Local Path-Based Integration: Applies a semi-local, path-based integration method to derive integrative conclusions about drug responses from the combined networks.
• Differential Drug Response Prediction: Compares integrated networks across conditions (e.g., estrogen receptor positive vs. negative breast cancer) to produce differential drug response predictions.
• Explainability: Identifies specific molecular interactions and changes that underlie high differential drug scores to enable interpretation of predicted responses.

Scientific Applications:

• Differential drug-response prediction in breast cancer: Applied to predict differential drug responses between patient groups such as ER-positive versus ER-negative breast cancer using transcriptomic, proteomic, phosphosite, and metabolomic data.
• Benchmarking and layer importance: Evaluated against cancer cell line ground truth and shown to outperform methods based solely on differential protein expression or PageRank, with proteomic and phosphosite layers identified as most informative.

Methodology:

Constructs condition-specific networks from each omics layer using correlation analysis, reduces and combines these networks into multi-layer heterogeneous molecular networks, applies a semi-local path-based integration, and compares integrated networks across conditions to predict differential drug responses.