Learning chemical sensitivity reveals mechanisms of cellular response

Area of Science:

• Computational biology
• Chemical biology
• Genomics

Background:

• Chemical probes are crucial for understanding molecular disease mechanisms.
• Comprehensive chemical screening across diverse biological models is challenging.
• Integrating genetic information with chemical structures can predict cellular responses.

Purpose of the Study:

• To develop an interpretable computational model, ChemProbe, for predicting cellular sensitivity to molecular probes and drugs.
• To leverage transcriptomics and chemical structures for in silico screening.
• To facilitate the study of molecular mechanisms underlying chemical sensitivity.

Main Methods:

• Developed ChemProbe, a machine learning model combining transcriptome data and chemical structures.
• Applied ChemProbe to predict chemical sensitivity in cancer cell lines and tumor samples.
• Analyzed model predictions to identify key transcriptome features and biological pathways.
• Validated predictions retrospectively for breast cancer treatment and prospectively in new cellular models.

Main Results:

• ChemProbe accurately predicts cellular sensitivity to a wide range of compounds.
• Model interpretation identified transcriptome features linked to compound targets and protein networks.
• Specific genes driving ferroptosis were identified through model analysis.
• Successful prospective validation in genetically modified cell lines demonstrated model robustness.

Conclusions:

• ChemProbe serves as an effective, interpretable in silico screening tool.
• The model facilitates research into the molecular basis of chemical sensitivity.
• ChemProbe aids in discovering potential therapeutic strategies and understanding drug mechanisms.