Fig. 4.

Comparison of ML-CDN2 and other network-based models. The bar graph shows the predictive performance of three models, which was estimated based on Pearson correlations (the number on the top of bars) between the predicted and observed IC₅₀ values. The first (Zhang) is Zhang et al.’s method. The second (Wei) is Wei et al.’s method. The third is the ML-CDN2 model from in this study using the CSN^path and DSNF