SVM-Fold: a tool for discriminative multi-class protein fold and superfamily recognition

BMC Bioinformatics

Table 2 Results for multi-class fold prediction in the remote homology detection set-up. Results for the adaptive codes method are reported for a SCOP benchmark data set (44 folds, 424 superfamilies, 809 families, with 381 test sequences).

Method (and optimization target)	Error	Balanced Error	Top 5 Error	Balanced Top 5 Error	Detection Rate at fdr = 1%	Detection Rate at fdr = 5%
PSI-BLAST	0.4094	0.4428	0.2966	0.3666	0.3123	0.3412
one-vs-all: Folds	0.3307	0.4565	0.1260	0.1954	0.2073	0.2493
one-vs-all: Folds, Sfams	0.3307	0.4565	0.1260	0.1954	0.2073	0.2336
Sigmoid Fitting: Folds	0.3465	0.4973	0.1706	0.3407	0.2178	0.2283
Adaptive Codes: Folds (zero-one)	0.3018	0.3769	0.1286	0.1862	0.2073	0.2362
Adaptive Codes: Folds (balanced)	0.3281	0.3766	0.1680	0.1702	0.2073	0.2362
Adaptive Codes: Folds, Sfams (zero-one)	0.2808	0.3749	0.1155	0.1770	0.2283	0.2493
Adaptive Codes: Folds, Sfams (balanced)	0.2887	0.3659	0.1260	0.1427	0.2231	0.2388
Adaptive Codes: Folds, Sfams, Fams (zero-one)	0.2493	0.3474	0.1024	0.1726	0.2231	0.2677
Adaptive Codes: Folds, Sfams, Fams (balanced)	0.2703	0.3445	0.1155	0.1418	0.2283	0.2703

ISSN: 1471-2105