LLM-PBC: Logic Learning Machine-Based Explainable Rules Accurately Stratify the Genetic Risk of Primary Biliary...

Area of Science:

• Genetics
• Computational Biology
• Medical Informatics

Background:

• Machine Learning (ML) application in genetic data analysis is emerging.
• Primary Biliary Cholangitis (PBC) risk prediction requires advanced methodologies.

Purpose of the Study:

• To evaluate the feasibility and accuracy of an ML model for PBC risk prediction.
• To apply ML to individual-level genetic data for disease risk assessment.

Main Methods:

• Utilized genome-wide significant variants from PBC GWAS in European ancestry.
• Applied ML, including Logic Learning Machine (LLM), to individual genomic data from Italian cohorts.
• Generated and validated predictive 'if-then' rules for PBC using genotype data.

Main Results:

• The best LLM model achieved 71.7% accuracy and 0.73 Area Under the Curve (AUC) in validation.
• Identified key genes (RIN3, KANSL1, TIMMDC1, TNPO3) in the highest-coverage predictive rule.
• Generated 38 intelligible rules for PBC prediction based on genetic variants.

Conclusions:

• This study is the first to apply ML to common variants for PBC prediction.
• The ML approach is computationally feasible and generates interpretable rules.
• The model can aid in predicting disease risk for individuals susceptible to PBC.