We established three distinct approaches to build prediction models [PMID: 37693565]. The first approach termed the 'baseline approach,' utilized clinical features (age, infancy vaccination, biological sex) and baseline task values as predictors for individual tasks. The second approach, 'MCIAbasic,' employed 10 multi-omics factors constructed using MCIA to predict individual tasks. Before applying MCIA, the harmonized datasets were further processed to impute missing data in the baseline training set using the Multiple Imputation by Chained Equations (MICE) algorithm. The MCIA's objective function aims to maximize the covariance between each individual omic dataset and a global data matrix composed of concatenated omic data blocks. Finally, the third approach, 'MCIAplus,' integrates the baseline approach and MCIAbasic. It uses clinical features, baseline task values, and the 10 MCIA factors identified through MCIAbasic as predictors for individual tasks.