Fig. 5
From: DNA methylation signatures of illicit drug injection and hepatitis C are associated with HIV frailty
A machine learning prediction of a penal of CpG sites on HIV frailty. a Prediction of 748 CpG sites on HIV frailty using a support vector machine (SVM) in the discovery sample. A set of 748 CpG sites was selected from epigenome-wide association analysis in 386 samples that was treated as a training set. A different set of samples (N = 238) selected from the same cohort was used as a test set. Receiver-operating characteristic (ROC) analysis shows the prediction of HIV frailty in 284 test samples. Left: area under the curve of the prediction of high HIV frailty (Veteran Aging Cohort Index, VACS index >50); right: area under the curve of the prediction of low HIV frailty (VACS index <16). b Permutation test (1000) of support vector machine-learning prediction for HIV high frailty (left) and low frailty (right)
