Skip to main content

Table 4 Cross validation experiments for predicting the enzymatic-reaction likeness.

From: KCF-S: KEGG Chemical Function and Substructure for improved interpretability and prediction in chemical bioinformatics

Chemical descriptors Vector dimension Diff-common L1SVM Diff-only L1SVM Baseline Random
   AUC AUPR AUC AUPR AUC AUPR AUC AUPR
KCF-S k3 53679 0.9841 0.2483 0.9827 0.1872 0.8254 0.0584 0.4981 0.0052
  10000 0.9839 0.2481 0.9824 0.1840 0.8299 0.0594 0.4985 0.0052
  1000 0.9814 0.2269 0.9773 0.1508 0.8397 0.0592 0.5006 0.0053
KCF-S k2 28152 0.9761 0.2144 0.9691 0.1330 0.8122 0.0503 0.4995 0.0053
  10000 0.9763 0.2148 0.9698 0.1366 0.8143 0.0501 0.4997 0.0053
  1000 0.9720 0.2012 0.9596 0.1029 0.8178 0.0481 0.4988 0.0053
KCF-S k1 11133 0.9702 0.1835 0.9620 0.1300 0.8184 0.0776 0.4962 0.0052
  10000 0.9699 0.1835 0.9600 0.1197 0.8187 0.0769 0.4960 0.0052
  1000 0.9676 0.1757 0.9475 0.0868 0.8208 0.0744 0.4963 0.0052
PubChem FP 879 0.9531 0.1341 0.9067 0.0571 0.8883 0.0667 0.5006 0.0052
MACCS FP 164 0.9275 0.0932 0.9097 0.0510 0.8200 0.0336 0.5001 0.0052
ATOM k3 99 0.9532 0.1362 0.9378 0.0703 0.8195 0.0492 0.4983 0.0052
BOND k3 973 0.9773 0.2023 0.9713 0.1319 0.8260 0.0546 0.5001 0.0053