ALACD

ALACD associates long noncoding RNAs (lncRNAs) with anti-cancer drugs using a bilevel optimization framework to predict drug-lncRNA associations and corresponding gene signatures.

Key Features:

• Bilevel Optimization Model: Implements a dual-layer optimization where the upper level optimizes gene signature overlap and the lower level imputes missing associations between lncRNAs and genes.
• Integration of Drug-Gene Information: Leverages existing small molecule target data to link drugs to gene signatures associated with lncRNAs.
• SVM Algorithm: Uses a Support Vector Machine to identify genes coexpressed with lncRNAs and to support enhancer-lncRNA-mRNA coexpression inference.
• Optimization Procedure (optimizing.R): Contains an R script for identifying optimal gene associations with lncRNAs and drugs via the bilevel optimization approach.
• Cross-Validation and Performance Evaluation (crossvalSVM.R, getperf.R): Includes scripts for SVM cross-validation and generation of evaluation criteria to assess predictive performance.
• Differential Expression Analysis (DEGanalysis.R): Provides an R script to assess differential expression levels of lncRNAs across cancer types.
• Survival Analysis (survival-analysis.R): Includes an R script to evaluate patient survival data for prognostic assessment of identified lncRNAs.
• Application to TCGA Cancer Types: Applied to 10 cancer types from The Cancer Genome Atlas (TCGA) with matched lncRNA and mRNA expression datasets.
• Functional and Molecular Pathway Analysis: Associates drug-linked gene signatures and lncRNAs with cancer development pathways.
• Prognostic Biomarker Potential: Supports identification of lncRNA-drug associations with evidence for prognostic relevance from survival analyses and literature.

Scientific Applications:

• Therapeutic target discovery: Identification of genes and lncRNAs that suggest potential targets for anti-cancer small molecules.
• Mechanistic interpretation: Elucidation of mechanisms of action by linking small molecule targets, gene signatures, and lncRNA coexpression.
• Cancer-type-specific association mapping: Discovery of lncRNA-drug associations specific to individual TCGA cancer types using matched lncRNA and mRNA data.
• Prognostic biomarker identification: Prioritization of lncRNAs with survival associations that may serve as prognostic markers or guide treatment stratification.

Methodology:

Uses a bilevel optimization framework (upper level optimiser for gene signature overlap; lower level imputes lncRNA–gene associations), integrates small molecule target data, applies an SVM for coexpression prediction, and employs R scripts including optimizing.R, crossvalSVM.R, getperf.R, DEGanalysis.R, and survival-analysis.R on matched TCGA lncRNA and mRNA expression datasets.