Designing combination therapies with modeling chaperoned machine learning

Area of Science:

• Computational biology
• Cancer research
• Pharmacology

Background:

• Chemotherapy resistance poses a significant obstacle to successful cancer treatment.
• Developing systematic methods to identify effective combination therapies is crucial for biomedical advancement.

Purpose of the Study:

• To create a computational model integrating biological knowledge and experimental data for rational combination therapy design.
• To predict the synergistic effects of cisplatin and TNF-related apoptosis-inducing ligand (TRAIL) on cancer cell death.

Main Methods:

• Developed a comprehensive computational model of cancer cell response to cisplatin and TRAIL.
• Validated model predictions using single-cell measurements of p53 dynamics, cell fate, and cell death.
• Applied diverse machine learning methods to analyze the validated model.

Main Results:

• The model accurately predicted that cisplatin and TRAIL combination therapy enhances cancer cell death with a "two-wave killing" pattern.
• Machine learning collectively identified molecular players (sensitizers) that enhance tumor cell sensitivity to cisplatin-induced apoptosis.
• Identified sensitizers align with existing experimental findings.

Conclusions:

• An experimentally validated mechanistic model, when analyzed with machine learning, can identify biologically relevant sensitizers.
• This approach facilitates the design of improved chemotherapy strategies to overcome drug resistance.
• The findings offer a pathway to enhance the efficacy of cancer treatments.