Statistical deconvolution of enthalpic energetic contributions to MHC-peptide binding affinity

BMC Structural Biology

Table 4 Best QSAR models for GFA and G/PLS based on full binding groove in comparison with the PLS method. *PLS Models based on Cerius GFA & G/PLS Models respectively.

Model	QSAR Analysis ¹Cerius ²SYBYL	Equation Length Terms + Constant	STATISTICAL PARAMETERS
			TRAINING SET									TEST SET
			LOF	SEP	r ²		(CV) r ²		SEE	PRESS		r		r ²		PRESS
Total + Electrostatic + VDW	GFA ¹	17	0.308		0.772		0.583			23.557		0.776		0.602		34.625
	G/PLS ¹	16			0.779		0.590			12.387		0.740		0.547		39.552
	PLS ² (GFA)	17		0.604		0.630		0.521	0.531		20.526		0.734		0.539		34.247
	PLS ² (G/PLS)	16		0.571		0.615		0.438	0.473		15.869		0.708		0.501		41.424
Electrostatic + VDW	GFA ¹	19	0.349		0.767		0.616			21.719		0.731		0.534		39.464
	G/PLS ¹	15			0.723		0.541			21.359		0.737		0.544		52.287
	PLS ² (GFA)	19		0.675		0.194		-0.055	0.611		25.038		0.613		0.376		66.115
	PLS ² (G/PLS)	15		0.558		0.669		0.526	0.466		15.863		0.803		0.645		48.045
Total	GFA ¹	18	0.200		0.861		0.751			11.071		0.781		0.610		51.105
	G/PLS ¹	15			0.665		0.299			39.635		0.796		0.634		40.751
	PLS ² (GFA)	18		0.654		0.496		0.350	0.576		23.550		0.777		0.604		58.579
	PLS ² (G/PLS)	15		0.583		0.697		0.537	0.471		16.669		0.861		0.742		35.160

ISSN: 1472-6807