Skip to main content
. 2012 Nov 23;6:139. doi: 10.1186/1752-153X-6-139

Table 7.

Comparison of virtual screening performance of SVM with those of other methods

Method Inhibitors in training set
Inhibitors in testing set
Virtual screening performance
No of inhibitors No of chemical families covered by inhibitors No of inhibitors No of chemical families covered by inhibitors Percent of inhibitors in chemical families covered by inhibitors in training set Yield No and Percent of identified true inhibitors outside training chemical families No and Percent of the 168K MDDR compounds identified as inhibitors No and Percent of the 9,305 MDDR compounds similar to the known inhibitors identified as virtual inhibitors
Support Vector Machines
1703
493
44
35
51.43%
70.45%
15(34.1%)
1,496 (0.89%)
719 (7.73%)
Tanimoto Similarity
36.84%
9(20.5%)
9,305 (5.54%)
9,305 (100%)
K Nearest Neighbour
38.64%
10(22.7%)
4,182 (2.49%)
1,169 (12.57%)
Probabilistic Neural Network           50.0% 13(29.5%) 4,386 (2.60%) 1,184 (12.72%)