Table 7. Performance for 1^st degree unclassifiable peptides, further split into 2^nd degree classifiable and unclassifiable peptides employing the simplified machine learning approach with understandable attributes (ML-simple) for prediction of IVIG binding^*.

	All 1st degreeunclassifiable	1st degree unclassifiable\2nd degree classifiable	1st degree unclassifiable\2nd degree unclassifiable
AUC	0.956	0.992	0.683
Accuracy	91.5%	97.8%	65.0%
Number of peptides	2,716	2,442	274

^*Comparison of AUC and accuracy when the classifier was 10-fold cross-validated on peptides of the 1^st degree unclassifiable set or on its peptide subsets that the first one classified correctly (2^nd degree classifiable) or incorrectly (2^nd degree unclassifiable), respectively. All classifiers used the interpretable attributes and logistic regression.