Research Paper Volume 16, Issue 5 pp 4075—4094
Genome-wide transcriptome profiling and development of age prediction models in the human brain
Figure 3. Age prediction model fit results in the training set and validation in the test set, developed with (A) Lasso, (B) EN, (C) XGBoost, and (D) LightGBM machine learning algorithms. (E) Results of the prediction by applying models developed using the training set to the external validation set. (F) Overlaps of genes whose expression levels were found to be important for the age prediction. Initially, 6,551 unique genes were considered, corresponding to the 9,296 probe sets showing linear relationship with age (FDR p-value < 0.05) in univariate linear regression modeling.