AI-driven multi-omics modeling of Myalgic Encephalomyelitis / Chronic Fatigue Syndrome, 2025, Xiong et al.

[ME/CFS Science Blog]

Referencing Fig. 2C, where scores are 1-mean squared error. It just stuck out to me because I’ve previously been part of analyses that tried to predict individual domain scores, on a much bigger cohort, using various supervised models, and only ever came close to those MSEs for maybe one score

Interesting thanks.

I wonder if they split the control group in group A and group B and do the same analysis on these groups (rather than ME/CFS versus healthy controls), how high the predictive power and AUC would be.