Deep gene compound profiler (DeepCOP)

Deep gene compound profiler (DeepCOP) predicts gene regulatory effects of low-molecular-weight compounds by integrating molecular fingerprint descriptors and gene ontology-derived gene descriptors with deep neural networks to identify differential gene regulation endpoints.

Key Features:

• Deep Learning Architecture: Utilizes deep neural networks trained on combined molecular fingerprint descriptors and gene descriptors derived from gene ontology terms.
• Data Integration: Integrates large-scale genomics and chemogenomics molecular data to enable prediction of differential gene regulation endpoints without requiring protein target interaction information.
• Input Data Sources: Operates on molecular fingerprint descriptors and gene ontology-derived gene descriptors with training labels from differential gene regulation endpoints in the LINCS database.
• Validation and Performance: Demonstrated robust performance with 10-fold cross-validation RAUC scores of 0.80 or higher and enrichment factors exceeding 5.
• External Validation Focus: Validated using an external RNA-Seq dataset examining the effects of three potent antiandrogens with distinct modes of action in LNCaP prostate cancer cells.

Scientific Applications:

• Drug discovery: Identification of potential small-molecule candidates that induce specific gene expression responses.
• Small-molecule regulator development: Acceleration of discovery of small-molecule regulators that influence cell development processes.
• Precision oncology: Facilitation of targeted therapeutic intervention research and candidate selection in cancer therapeutics, including prostate cancer contexts.

Methodology:

Deep neural networks were trained on combined molecular fingerprint descriptors and gene ontology-derived gene descriptors using differential gene regulation endpoints from the LINCS database; model performance was assessed by 10-fold cross-validation (RAUC and enrichment factors) and validated with an external RNA-Seq dataset focused on three potent antiandrogens in LNCaP prostate cancer cells.