TRIOMPHE: Transcriptome-Based Inference and Generation of Molecules with Desired Phenotypes by Machine Learning.

Area of Science:

• Computational chemistry
• Pharmacology
• Genomics

Background:

• Efficient identification of small molecules with desired phenotypes is a major challenge in drug discovery.
• Understanding cellular responses to compound treatment and genetic perturbations is crucial for ligand-target interactions.

Purpose of the Study:

• To propose a novel computational method, TRIOMPHE (transcriptome-based inference and generation of molecules with desired phenotypes), for omics-based de novo drug design.
• To develop machine learning models for generating new molecules with desired transcriptome profiles.

Main Methods:

• Investigated correlations between chemically and genetically induced transcriptome profiles.
• Developed variational autoencoder-based machine learning methods for molecule generation.
• Utilized desired transcriptome profiles to guide the design of molecules with specific bioactivities.

Main Results:

• Generated chemically valid molecules with predicted biological activities for 10 target proteins.
• TRIOMPHE outperformed previous methods for de novo drug design with the same objective.
• Demonstrated the utility of transcriptome profiles in designing molecules for targeted drug discovery.

Conclusions:

• TRIOMPHE offers a novel computational approach for omics-based de novo drug design.
• The method facilitates the automatic design of molecules likely to exhibit desired bioactivities against target proteins.
• This omics-based structure generator is a valuable tool for developing drugs against various targets.