Skip to main content

Table 7 Atom-only representation 10-folds cross-validation predictive accuracies for ProGolem using different recall selection methods

From: Automated identification of protein-ligand interaction features using Inductive Logic Programming: a hexose binding case study

  Recall selection method
Fold Primary sequence Randomized Domain-dependent
1 43.8% 56.3% 87.5%
2 62.5% 93.8% 78.5%
3 81.3% 87.5% 87.5%
4 56.3% 50.0% 43.8%
5 68.8% 68.8% 81.3%
6 37.5% 56.3% 81.3%
7 56.3% 62.5% 75.0%
8 68.8% 68.8% 81.3%
9 62.5% 81.3% 62.5%
10 56.3% 62.5% 68.8%
Mean 59.4% 68.8% 74.8%
Std Dev 12.6% 14.4% 13.4%