Figure 2.

Performance of tissue functional gene networks. (A) Network performance of each network based on 5-fold cross-validation. The cross-validation is conducted based on gene holdout rather than edge holdout; that is, genes are split into 5-folds to make sure that any gene used for training models during cross-validation will not occur in the holdout set. (B) Comparison of XGBoost with Bayesian classifiers in constructing tissue functional gene networks.