A computational method for designing diverse linear epitopes including citrullinated peptides with desired binding affinities to intravenous immunoglobulin

Table 1 Performance of the various classifiers used within the Pythia method

Features	AUROC	AUPR	ΔAUROC	ΔAUPR
k-spectrum	0.85	0.70	−0.043	−0.072
Sparse Spatial Sample	0.87	0.73	−0.023	−0.042
Nonlinear Fisher Mat.	0.86	0.69	−0.024	−0.082
Statistical Analysis Mat.	0.85	0.67	−0.025	−0.102
BLOSUM Encoding	0.86	0.70	−0.024	−0.072
Local Composition^a	0.88	0.74	−0.013	−0.032
Structure	0.74	0.53	−0.153	−0.242
Ensemble	0.89	0.77

ISSN: 1471-2105