Cancer Mutations Converge on a Collection of Protein Assemblies to Predict Resistance to Replication Stress

Area of Science:

• Oncology
• Computational Biology
• Genetics

Background:

• Cancer cells' rapid proliferation makes them sensitive to therapies causing DNA replication stress (RS).
• However, many tumors develop resistance through poorly understood molecular pathways.
• Identifying these pathways is crucial for developing effective cancer treatments.

Purpose of the Study:

• To develop predictive models elucidating how cancer mutations impact response to common replication stress-inducing (RSi) agents.
• To interpret the mechanistic basis of drug response and resistance.
• To identify molecular assemblies and genetic alterations that govern drug sensitivity.

Main Methods:

• Ensemble of deep learning predictive models for multidrug response prediction.
• Integration of genetic alterations across hundreds of genes.
• Loss-of-function genetic screens to validate identified molecular assemblies.
• Application of models to patient data (cisplatin-treated cervical cancer).

Main Results:

• Identified 41 molecular assemblies integrating hundreds of gene alterations for accurate drug response prediction.
• 30 of these assemblies were shown by genetic screens to regulate drug sensitivity or replication restart.
• Highlighted an RTK-JAK-STAT assembly as a key mechanism of resistance in cisplatin-treated cervical cancer patients.

Conclusions:

• Defined a compendium of mechanisms by which cancer mutations affect therapeutic responses.
• Demonstrated the utility of deep learning for predicting drug response and interpreting resistance mechanisms.
• Findings have significant implications for advancing precision medicine in oncology.