Supplementary Figure 3: Comparison of prediction models in data from the UK Biobank.
From: Common brain disorders are associated with heritable patterns of apparent aging of the brain
Our main analysis used xgboost models for estimating brain age. The figure compares predicted brain age (a) and brain age gaps (b) from xgboost to shrinkage linear models (slm) and to a free public brain age estimation approach (brainageR). Predicted brain age and resulting brain age gaps were highly correlated between approaches (Pearson r; analysis performed in n=23,602 healthy individuals from the UK Biobank; two-sided; all P < FDR; Benjamini-Hochberg). All brain age gaps account for age, age², sex, scanning site and Euler number.
