Bayesian latent trait modeling of migraine symptom data

Area of Science:

• Genetics
• Neurology
• Biostatistics

Background:

• Accurate disease phenotyping is crucial for identifying genetic causes of complex conditions like migraine.
• Current migraine diagnosis relies on International Headache Society criteria, which were previously analyzed using latent class analysis (LCA) revealing a four-class model.
• The ordered nature of migraine symptom classes suggests a continuous severity variable might be a more effective model.

Purpose of the Study:

• To compare item response theory and latent class analysis (LCA) within a Bayesian framework for migraine phenotyping.
• To develop and validate continuous measures of migraine severity for genetic studies.
• To identify genetic loci associated with migraine susceptibility.

Main Methods:

• Bayesian modeling comparing item response theory and LCA.
• Deviance information criterion for model fit assessment.
• Genome-wide linkage analysis using Merlin-qtl on a phenotyped population sample.

Main Results:

• Latent class analysis (LCA) with four classes was again preferred.
• Phenotypic trait values from both models were highly correlated (0.99) after transformation.
• Estimated heritability was 0.37, with significant linkage to chromosome 7q31-q33 and suggestive linkage to chromosomes 1 and 2.

Conclusions:

• Continuous measures of disease severity are powerful tools for identifying genes contributing to migraine susceptibility.
• The developed phenotyping models provide a robust approach for future genetic research in migraine.
• Genetic linkage findings highlight specific chromosomal regions for further investigation into migraine's genetic architecture.