In a new study, researchers argue they established a well-performing risk score model for multiple sclerosis based on genetic burden alone, demonstrating the potential usefulness of phenotype-based risk assessment. They suggest the combination risk models have the potential to aid in early screening and diagnosis.

MS is a demyelinating disease whose likelihood of development is derived from a combination of genetic and environmental risk factors. While there is no cure for the disease, many disease-modifying therapies have demonstrated the ability to slow or halt progression. Current research indicates improved long-term health outcomes for patients are linked to earlier diagnosis and treatment initiation. 

Researchers from Brigham Young University and the National Institutes of Health set out to develop a combination genotype-phenotype risk prediction model. It is based on possession of known MS genetic risk variants and presence of key International Classification of Diseases billing codes in patients’ electronic health records. 

Participants were recruited from within the All of Us research platform. Cases were defined as individuals possessing two or more ICD codes for MS that were separated by at least one day, and included individuals with African, Admixed American, and European genetic ancestry. Controls included any participant with no ICD codes for both MS and neuromyelitis optica. Genetic and polygenic risk scoring using established and potential variants was performed and evaluated for predictive power through receiver-operating characteristic curve and area-under-the-curve analysis. Participants were then divided into training and testing sets, and a phenome-wide association study was performed using PheTK to identify phecodes significantly linked to MS. They then developed a phenotype risk score model and measured its performance. A joint genetic risk score/phenotype risk score model was evaluated using receiver-operating characteristic/area-under-the-curve analysis to determine its predictive power. 

The genetic risk score outperformed the phenotype risk score in all groups with predictive area-under-the-curve ranging from .57-.70 while the polygenic risk scoring reached .54. The European ancestry genetic risk score performed the best while the African ancestry genetic risk score was the least performant. The highest performing genetic risk score model utilized only established risk variants with the addition of a proxy SNP for HLA-DRB*15:01. 

The phenome-wide association studies identified 122 phecodes significantly linked to MS. Subsequent phenotype risk scores demonstrated strong predictive power, with an area-under-the-curve of .84 in the general cohort, and scores differed significantly for cases and controls both broadly and when stratified by predicted genetic ancestry. 

Combining the phenotype risk score with the best performing genetic risk score model strengthened the general cohort area-under-the-curve to .867. The combination of phenotype risk score/genetic risk score also improved predictive values when stratified by predicted genetic ancestry. 

The researchers argue they established a well-performing risk score model for MS based on genetic burden alone, and demonstrate the potential utility of phenotype-based risk scoring. They highlighted the disparities across genetic ancestries along with the need for research in underserved populations. Combination genotype-phenotype risk models have the potential to aid in early screening and diagnosis of MS.

The study was presented at ACTRIMS 2025.