Logic models to predict continuous outputs based on binary inputs with an application to personalized cancer therapy.

Area of Science:

• Computational biology
• Machine learning
• Genomics

Background:

• Machine learning models from large datasets are often uninterpretable.
• This lack of interpretability hinders hypothesis generation for experimental testing.

Purpose of the Study:

• To present Logic Optimization for Binary Input to Continuous Output (LOBICO), a novel computational approach.
• To infer small, interpretable logic models explaining continuous output variables from binary input features.

Main Methods:

• Applied LOBICO to a large cancer cell line panel.
• Utilized continuous information for robust and accurate logic model inference.
• Implemented ability to uncover logic models around predefined operating points (sensitivity/specificity).

Main Results:

• Logic combinations of multiple mutations are more predictive of drug response than single gene predictors.
• Continuous information integration yields more robust and accurate logic models.
• LOBICO successfully identified interpretable logic models from complex biological data.

Conclusions:

• LOBICO provides an interpretable alternative to complex machine learning models in biology.
• The approach facilitates hypothesis generation for experimental validation in cancer research.
• LOBICO represents a significant advancement towards the practical application of interpretable logic models.