Statistical method uses genetic data to estimate probabilities across four psychiatric diagnoses
Researchers published a statistical framework in Nature Genetics called differential diagnosis polygenic risk score that estimates probabilities across schizophrenia, bipolar disorder, major depressive disorder, and controls. The method models variance-covariance structure across disorders using polygenic risk scores and prior clinical probabilities. The team used Psychiatric Genomics Consortium data from three genome-wide association studies with 41,917 to 173,140 cases per disorder, totaling 1,048,683 individuals in training sets. They tested the model on 11,460 individuals split equally across the four categories. The model was well calibrated, and true diagnosis probabilities in the top deciles of predicted probabilities exceeded baseline probabilities. Results matched methods requiring tuning data.
The framework offers a way to model genetic risk across overlapping psychiatric conditions simultaneously rather than using separate case-control comparisons, though clinical validation in prospective settings has not yet been done.
Written by the Genomes desk from the primary source linked above and checked against it. Research use only; not medical advice. Corrections: [email protected].