Skip to main content
. Author manuscript; available in PMC: 2021 Sep 1.
Published in final edited form as: Mach Learn Sci Technol. 2021 May 13;2(3):035015. doi: 10.1088/2632-2153/abe6d6

Figure 3.

Figure 3.

Fine-tuned results of ratio and distance threshold parameters in GCNN model. (A) The ratio between number of negative labels and number of positive labels was fine-tuned. Ratio of 2 is considered reaching a balance between recall and precision. (B) The atomic distance threshold was fine-tuned from 7 to 11 Å. There is no significant increase in F1 score after 10Å, which is selected as the distance cutoff. For each parameter value, GCNN was run 10 times independently.