From multi-omics data to the cancer druggable gene discovery: a novel machine learning-based approach

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Precision medicine and targeted therapies in cancer treatment rely on identifying druggable genes.
• The complexity and diversity of multi-omics data present challenges in discovering rare cancer-druggable genes.

Purpose of the Study:

• To propose a novel machine learning-based method, Deep Forest for Cancer Druggable Genes Discovery (DF-CAGE), for identifying cancer-druggable genes.
• To analyze the landscape of cancer-druggable genes using integrated multi-omics data from The Cancer Genome Atlas (TCGA).

Main Methods:

• DF-CAGE integrated somatic mutations, copy number variants (CNVs), DNA methylation, and RNA-Seq data from approximately 10,000 TCGA profiles.
• The method was validated against established datasets like OncoKB, Target, and DrugBank.
• Analysis of omics data contribution to druggable gene identification was performed.

Main Results:

• DF-CAGE identified 465 potential cancer-druggable genes among approximately 20,000 protein-coding genes.
• The identified candidate druggable genes (CDGs) were found to be clinically meaningful and categorized into known, reliable, and potential sets.
• DF-CAGE demonstrated excellent performance, discovering commonalities among known cancer-druggable genes through multi-omics data integration.

Conclusions:

• DF-CAGE effectively identifies potential cancer-druggable genes by leveraging multi-omics data, advancing precision cancer treatment.
• The study highlights the clinical significance of identified candidate druggable genes.
• Copy number variations (CNVs), gene rearrangements, and mutation rates were found to be key contributors to druggable gene identification by DF-CAGE, guiding future drug development.