Machine learning integration for predicting the effect of single amino acid substitutions on protein stability

BMC Structural Biology

Table 25 Comparison of our results for S1615 data set with previously published studies.

Ref.	Method	Data Set Size	Accuracy	Information
[41]	SVM	2048	0.77 (20-fold cv)	Seq
[42]	SVM	1383 *	0.73 (20-fold cv)	Seq
[9]	NN NN+FOLDX	1615	0.79 (20-fold cv) 0.87 (test set†) 0.93 (test set†)	Seq+Str
[2]	SVM	1496‡	SO: 0.84, TO: 0.85, ST: 0.85 (20-fold cv) SO: 0.86, TO: 0.86, ST: 0.86 (test set)	Seq+Str
[31]	iPTREE	1615	0.87 (10-fold cv)	Seq+Str
Ours	Early Late Intermediate	1122 (training) 383 (test)	0.842 (20-fold cv), 0.904 (test set) 0.847 (20-fold cv), 0.903 (test set) 0.826 (20-fold cv), 0.879 (test set)	Seq+Str

*Filtered from the set of 2048 mutations [41].
† A subset of the training set that was previously used in training.
‡ Filtered from the set of 1615 mutations [9].
Machine learning method, data set, performance assessment are the main features to be compared. (Seq: Sequence-based information, Seq+Str: Sequence- and structure-based information)

ISSN: 1472-6807