Inexpensive, non-invasive biomarkers predict Alzheimer transition using machine learning analysis of the Alzheimer’s Disease Neuroimaging (ADNI) database
Fig 4
RF Classifier with subset of plasma BMs, compressed cognitive scores, and demographics.
The features include the 2 compressed ADNI cognitive variables, 14 plasma analytes chosen from the literature (including 9 Apolipoproteins (Ai, Aii, Aiv, B, Ci, Ciii, D, E, H) plus leptin, insulin, CRP, vitronectin). Demographics included, systolic BP, BMI, age, gender. The importance is displayed as the GINI coefficient below in Table 1. AUC = 0.71; PR curve = 0.60.