Personalised Medicine for Colorectal Cancer Using Mechanism-Based Machine Learning Models

Area of Science:

• Oncology
• Systems Biology
• Bioinformatics

Background:

• Signal transduction networks (STNs) are crucial for understanding cancer mechanisms.
• Patient-specific mathematical models can improve cancer stratification and identify drug targets.
• The WNT pathway is a critical STN in colorectal cancer.

Purpose of the Study:

• To identify critical genes and proteins in the WNT pathway associated with event-free survival in colorectal cancer.
• To develop mechanism-based machine learning models for predicting patient-specific STN activity.
• To improve patient stratification and identify novel therapeutic targets.

Main Methods:

• Utilized mechanism-based machine learning models on colorectal cancer data.
• Integrated data from The Cancer Genome Atlas and Clinical Proteomic Tumour Analysis Consortium.
• Computed patient-specific STN activity scores using PROGENy.
• Built linear regression models using existing and novel WNT pathway genes and proteins.

Main Results:

• Identified the WNT pathway as significantly associated with event-free survival.
• Models incorporating novel, event-free survival-associated genes demonstrated superior predictive power.
• Identified specific genes (e.g., DVL3, FZD5, RAC1) for future WNT pathway models.

Conclusions:

• Mechanism-based machine learning effectively identifies novel genes and proteins impacting STN heterogeneity.
• This approach enhances understanding of patient-specific differences in STN activity and clinical outcomes.
• The findings facilitate improved patient stratification and the discovery of potential drug targets in colorectal cancer.