Statistical modeling to quantify the uncertainty of FoldX-predicted protein folding and binding stability

BMC Bioinformatics

Table 4 Comparison of model performances

	Model 1	Model 2	Model 3	Model 4	Model 5 (full)
Folding dataset:
Coverage (%)	94.5	94.5	94.5	94.3	94.6
Median width (kcal/mol)	3.78	3.75	2.88	2.84	2.89
Adjusted R²	0.17	0.22	0.42	0.45	0.44
BIC	2279	2284	1928	1936	1930
Binding dataset:
Coverage (%)	94.6	95.7	95.1	94.6	94.8
Median width (kcal/mol)	3.81	3.74	3.52	3.51	3.47
Adjusted R²	0.62	0.66	0.52	0.56	0.57
BIC	1814	1813	1710	1709	1700

Leave-one-system-out cross validation results show all models achieve approximately 95% coverage. Median widths decreased in Models 3–5 which used MD snapshots data. While R² values were referenced during evaluation, more emphasis was given to the cross validation results (coverage and median width.)

ISSN: 1471-2105