does not want to lose any potential active compound/drug candidate particularly the
one like the mispredicted compound of the test set (compound no. 13) which has a
high MPS value (MPS = 83). Clearly, a number of active compounds have got high
MPS values including compound no. 8 which represents a potent anti-HIV drug—
Didanosine—and is a test set compound (Table 5). The method has also produced
high MPS values for a number of training set active compounds too like compound
nos. 5, 7, 18, 19 (Table 5). Therefore, picking at least a couple of top scoring (from
MPS values) compounds out of them from prioritization point of view may help
screen useful drug candidates using the present method. This finding therefore
indicates that this method can be used for creating suitable splits in getting a
reasonably useful training set from an available data set and help screen putative
active compounds for drug discovery.
Table 5 Assigned and predicted activities using D
−4 index and Molecular Priority Score
(MPS) of 20 nucleoside analogues divided into 14 training set and 6 test set compounds
Sr. no.
Compound no. #
Activity
a
MPS
b
Assigned
Predicted
Value
Training set
1
4
+
+
6 5
2
5
+
+
8 3
3
6
+
+
8
4
7
+
+
103
5
9
+
+
5 5
6
1 8
+
+
9 7
7
1 9
+
+
9 8
8
1
–
–
−56
9
2
–
–
−36
10
3
–
–
−48
11
10
–
+
8
12
14
–
–
−13
13
15
–
–
−36
14
16
–
–
−48
Test set
1
8
+
+
6 5
2
1 2
+
+
6 5
3
2 0
+
+
5 0
4
1 1
–
–
−48
5
1 3
–
+
8 3
6
1 7
–
–
−6
a (+) means active, (−) means inactive and (#) means incorrect prediction
b
The details for the computation of MPS value are described in methods section
#Compound numbers are correspond to those in Table 4
96
Md.I. H. Rizvi et al.