Table I.

Summary of Data Sets and the Results of Cross-Validation Tests

		Number of sites		AUC		sn at sp = 0.90		sn at sp = 0.95		sn at sp = 0.99
Modification type	Residue	Positive	Negative	No PSSM	PSSM	No PSSM	PSSM	No PSSM	PSSM	No PSSM	PSSM
Acetylation	K	6,848	149,314	0.688	0.713	0.277	0.312	0.168	0.188	0.046	0.057
ADP-ribosylation	E, R	108	4,681	0.739	0.753	0.356	0.369	0.221	0.236	0.063	0.062
Amidation	All	457	29,966*	0.964	0.967	0.923	0.930	0.851	0.866	0.570	0.615
C-linked glycosylation	W	32	118	0.938	0.928	0.756	0.837	0.606	0.750	0.454	0.415
Carboxylation	E	112	1,063	0.920	0.939	0.795	0.843	0.641	0.767	0.359	0.535
Disulfide linkage	C	9,736	7,101	0.646	0.783	0.182	0.391	0.110	0.246	0.037	0.078
Farnesylation	C	41	59*	0.857	0.862	0.533	0.633	0.319	0.225	0.174	0.041
Geranylgeranylation	C	30	43*	0.866	0.919	0.571	0.687	0.393	0.596	0.230	0.534
GPI-anchor amidation	N	84	2,362	0.961	0.966	0.908	0.905	0.841	0.853	0.518	0.528
Hydroxylation	K, P, Y	219	4,209	0.832	0.907	0.535	0.732	0.388	0.586	0.109	0.253
Methylation	K, R	628	18,561	0.660	0.674	0.319	0.349	0.243	0.264	0.130	0.127
Myristoylation	G	99	119*	0.792	0.852	0.353	0.514	0.175	0.354	0.038	0.008
N-linked glycosylation	N	11,286	78,050	0.790	0.806	0.215	0.330	0.066	0.160	0.018	0.030
N-terminal acetylation	A, G, M, S, T	1,310	2,002*	0.821	0.836	0.471	0.503	0.310	0.331	0.093	0.106
O-linked glycosylation	S, T	1,427	44,048	0.731	0.749	0.350	0.376	0.228	0.253	0.059	0.082
Palmitoylation	C	245	1,298	0.856	0.881	0.625	0.679	0.467	0.525	0.192	0.244
Phosphorylation	S, T, Y	90,058	320,506	0.771	0.777	0.422	0.437	0.296	0.312	0.113	0.116
Proteolytic cleavage	All	997	257,783	0.727	0.759	0.379	0.420	0.264	0.291	0.085	0.102
PUPylation	K	87	1,077	0.658	0.786	0.218	0.436	0.123	0.256	0.042	0.059
Pyrrolidone carb. acid	Q	275	2,789	0.880	0.906	0.682	0.770	0.538	0.658	0.188	0.389
Sulfation	Y	121	667	0.913	0.930	0.772	0.832	0.575	0.629	0.304	0.268
SUMOylation	K	744	17,539	0.742	0.739	0.419	0.458	0.311	0.360	0.135	0.172
Ubiquitylation	K	1,092	27,774	0.583	0.605	0.164	0.185	0.089	0.101	0.020	0.025

Each row shows mean performance for one PTM type, combined over all amino acids and subdata sets (motifs and non-motifs). A breakdown of performance for each amino acid and data set is provided in Supporting Information, Table S1. The “No PSSM” column represents the basic classification model and the “PSSM” column represents the model enhanced with evolutionary features. Area under the ROC curve (AUC) is shown for each PTM as well as sensitivity (sn; true positive rate) at different levels of specificity (sp; true negative rate). ROC curves for each amino acid and data set are provided in Supporting Information, Figure S3. Values marked in bold indicate the better-performing model. The data sets marked with a “^*” indicate that the negatives were obtained from proteins different to those containing the positives through a random sampling procedure (See Materials and Methods).