DRIM

DRIM integrates multi-omics (genomics, transcriptomics) and time-series gene expression data to identify IC50- or AUC-associated mediator genes, perturbed sub-pathways, and regulatory networks underlying drug response.

Key Features:

• Multi-Omics Data Integration: Integrates genomics and transcriptomics with drug sensitivity metrics such as IC50 and AUC to link molecular variation to drug response.
• Time-Series Analysis: Incorporates time-series gene expression data to construct dynamic models of pathway perturbation following drug treatment.
• Pathway and Sub-Pathway Analysis: Embeds multi-omics data into a gene-centric vector space and identifies perturbed sub-pathways affected by drugs.
• Embedding and Dimensionality Reduction: Applies tensor decomposition and autoencoder deep learning models for embedding multi-omics data into gene-centric representations.
• Identification of Mediator Genes: Determines potential mediator genes associated with IC50-related multi-omics signals.
• Network Construction and Influence Tracing: Constructs networks of transcription factors (TFs), mediator genes, and perturbed pathways and employs influence maximization to trace paths from TFs to perturbed pathways.

Scientific Applications:

• Pharmacogenomics: Elucidates molecular determinants of variable drug response by linking multi-omics alterations to IC50/AUC measures.
• Personalized Medicine: Identifies candidate mediator genes and perturbed sub-pathways for tailoring therapeutic strategies to molecular profiles.
• Cell Line–Specific Response Analysis: Analyzes differential drug responses across cell lines, including applications to breast cancer cell lines with varying drug sensitivities.

Methodology:

Embedding multi-omics into a gene-centric vector space using tensor decomposition and autoencoder models; associating multi-omics with IC50/AUC to determine potential mediator genes; perturbed pathway and sub-pathway analysis; constructing time-varying perturbed sub-pathways and networks from time-series gene expression; applying influence maximization to trace TF-to-pathway paths.