Mining Complex Genetic Patterns Conferring Multiple Sclerosis Risk.

Area of Science:

• Genetics
• Computational Biology
• Neuroimmunology

Background:

• Complex genetic interactions, including gene-gene and gene-environment effects, contribute significantly to multiple sclerosis (MS) heritability.
• Machine learning and data mining offer promising avenues for exploring these higher-order genetic relationships in MS, though their application has been limited.

Purpose of the Study:

• To apply association rule mining (ARM), a machine learning technique, to identify patterns among known MS risk variants.
• To uncover specific combinations of genetic variants that confer disproportionately elevated risk for developing MS.

Main Methods:

• Association rule mining (ARM) was employed on genetic data from non-Latinx MS cases and controls.
• The analysis focused on known MS risk variants, including HLA variants and common autosomal variants.
• Probabilistic measures (confidence and support) were utilized to mine significant association rules.

Main Results:

• 114 association rules met the minimum thresholds for confidence (80%) and support (5%).
• The top-ranking rule identified a combination of HLA-DRB1*15:01, SLC30A7-rs56678847, and AC093277.1-rs6880809, significantly increasing MS risk (OR=20.2).
• INTS8-rs78727559 was the most frequently shared variant across identified rules (32.5%).

Conclusions:

• This study demonstrates that specific combinations of MS risk variants can disproportionately elevate an individual's risk for the disease.
• The application of a robust analytical framework, ARM, successfully identified these complex genetic patterns in a moderately sized cohort.
• Findings highlight the importance of considering combinatorial genetic effects in understanding MS etiology.