Exploring the potential of 3D Zernike descriptors and SVM for protein–protein interface prediction

BMC Bioinformatics

Table 7 The selected (best) SVM model for each protein class, i.e. the penalty C, the kernel function and its parameters (γ, d, r)

Protein complex class	Receptor ligand	Bound unbound	No. features	kernel function	C	γ	d	r
A	r	b	109	sigmoid	495.33	0.00054	N/A	1.44470
		u	96	rbf	1365.14	0.00039	N/A	N/A
	l	b	89	linear	46.05	N/A	N/A	N/A
		u	85	linear	221.64	N/A	N/A	N/A
AB	r	b	117	poly	23.87	0.03006	2	1.73464
		u	108	poly	426.47	0.01110	3	0.01539
	l	b	78	poly	2157.88	0.01906	7	0.17614
		u	75	poly	4362.45	0.03470	10	-0.03613
EI	r	b	105	poly	1514.50	0.00003	3	-0.15922
		u	129	sigmoid	33.32	0.00029	N/A	-1.61953
	l	b	91	sigmoid	213.15	0.00065	N/A	0.47294
		u	80	poly	1916.02	0.01531	4	0.13840
ER	r	b	115	rbf	9.22	0.00366	N/A	N/A
		u	126	rbf	298.47	0.00222	N/A	N/A
	l	b	100	sigmoid	157.32	0.00024	N/A	-0.24272
		u	100	poly	1001.44	0.00597	5	0.00039
ES	r	b	84	linear	196.85	N/A	N/A	N/A
		u	79	linear	7010.36	N/A	N/A	N/A
	l	b	83	poly	954.76	0.00581	6	1.00104
		u	86	poly	721.43	0.02692	6	0.00022
OG	r	b	102	poly	8543.28	0.01682	6	0.00004
		u	107	rbf	12.42	0.00062	N/A	N/A
	l	b	92	poly	257.51	0.00575	3	-0.00191
		u	78	poly	3421.90	0.01659	8	0.00014
OR	r	b	97	linear	281.56	N/A	N/A	N/A
		u	68	linear	1804.59	N/A	N/A	N/A
	l	b	79	poly	5502.26	0.01908	9	0.00113
		u	100	sigmoid	63.94	0.00261	N/A	-1.90377
OX	r	b	141	rbf	60.29	0.00029	N/A	N/A
		u	122	rbf	747.39	0.00006	N/A	N/A
	l	b	132	poly	383.62	0.02146	8	0.04259
		u	118	poly	779.96	0.02933	9	0.05214
Generic model		b	83	sigmoid	148.639	0.02312	N/A	-1.44779
		u	76	sigmoid	3218.238	0.00196	N/A	1.92731

The “No. features” column indicates the number of selected features resulting from the Randomized Logistic Regression algorithm

ISSN: 1471-2105