Skip to main content
. 2021 Apr 1;11:7400. doi: 10.1038/s41598-021-86921-9

Figure 4.

Figure 4

Heatmap showing model performance using one data set as the training set and another data set as the testing set. Model performance is represented by the negative log10 p-value of the correlation between the predictions versus the ground truth data. Gray indicates that the model is not available; i.e., the drug does not exist in either data set. Models with a more intense green colors have better performance.