Prediction of new-onset migraine using clinical-genotypic data from the HUNT Study: a machine learning analysis

Area of Science:

• Genetics and bioinformatics
• Epidemiology
• Machine learning in healthcare

Background:

• Migraine has a significant genetic component, but identified risk loci explain only a fraction of its heritability.
• The
• missing heritability
• highlights the need for novel approaches to understand migraine risk.
• Investigating genetic and clinical interactions is crucial for predicting migraine onset.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting new-onset migraine.
• To assess the predictive power of combining genetic data with clinical features.
• To identify key genetic and clinical factors associated with migraine development.

Main Methods:

• Population-based cohort study using data from the Trøndelag Health Study (HUNT2 and HUNT3).
• Genome-wide genotyping and validated migraine questionnaires based on International Classification of Headache Disorders (ICHD) criteria.
• Development and optimization of machine learning models, including decision-tree classifiers, utilizing genetic variants and clinical variables.

Main Results:

• The best model, combining genetic and clinical data, achieved an Area Under the Receiver Operating Characteristics curve (AUC) of 0.72.
• Models using only clinical data (AUC 0.68) outperformed those using only genetic data (AUC 0.56).
• Key predictors included age, marital status, work situation, and specific genetic variants.

Conclusions:

• Combining genotype with demographic and non-headache clinical data can predict new-onset migraine in about two-thirds of cases.
• The findings suggest significant genotypic-phenotypic interactions influence migraine onset.
• This approach offers a promising tool for identifying individuals at risk of developing migraine.