Table 2. Training datasets for different classifiers.

	Problem 1	Problem 2	Problem 3
Beneficial	208	878 = 208(B)+670(N)	208
Neutral	-		670
Detrimental	1,076	1,076	1,076
Unlabeled	17,692	17,692	17,692

Different combinations of three types of nsSNPs are used for each of the three classification problems. The set of 878 PPI preserving mutations included both beneficial (B) and neutral (N) nsSNPs. The unlabeled set was used solely for semi-supervised learning methods.