Table 6.

Accuracy (ACC) and area under the precision-recall curve (AUPRC) of the top five one-feature and two-feature combinations of the logistic regression (LR), support vector machine (SVM), k-nearest neighbors (KNN) and decision tree (DT) models for classifying low/high viscosity. There are 20 mAbs in this study. The ACC and AUPRC are averaged from 100 randomly generated 4-fold cross-validation sets. The baseline ACC is 0.70 and the baseline AUPRC is 0.30

	One-feature	ACC	AUPRC	Two-features		ACC	AUPRC
	N_neg_VH	0.79	0.57	SCM_neg_VH	SCM_neg_VL	0.86	0.70
	SCM_neg_VL	0.77	0.54	N_neg_VH	SCM_neg_VL	0.84	0.68
LR	net charges_VH	0.78	0.53	N_neg_VH	net charges_VL	0.83	0.67
	N_neg_VL	0.77	0.51	SCM_neg_VL	SCM_pos_VH	0.83	0.66
	net charges_VL	0.74	0.48	net charges_VH	net charges_VL	0.81	0.65
	N_neg_VH	0.76	0.47	N_philic_VH	SAP_pos_VL	0.82	0.64
	net charges_VH	0.74	0.46	N_philic_Fv	SAP_pos_VL	0.82	0.63
SVM	SCM_neg_VL	0.72	0.45	N_philic_Fv	N_neg_VH	0.82	0.60
	mAbCSP	0.74	0.37	N_phobic_VL	N_neg_VH	0.82	0.60
	N_neg_VL	0.70	0.34	N_philic_VH	N_neg_VH	0.81	0.58
	HVI	0.76	0.59	N_pos_VL	N_neg_VH	0.83	0.66
	SAP_pos_VL	0.82	0.65	N_philic_Fv	FvCSP	0.82	0.64
KNN	SCM_neg_VL	0.74	0.52	N_pos_VL	net charges_VH	0.83	0.64
	net charges_VH	0.78	0.51	SCM_neg_VH	SCM_neg_VL	0.82	0.62
	N_neg_VH	0.78	0.5	N_philic_VH	FvCSP	0.76	0.62
	SAP_pos_VL	0.73	0.52	N_phobic_VL	SAP_pos_VL	0.84	0.74
	net charges_VH	0.77	0.51	N_neg_Fv	SCM_pos_VL	0.78	0.60
DT	N_neg_mAb	0.79	0.51	N_neg_mAb	net charges_VL	0.77	0.60
	SCM_pos_VL	0.75	0.49	N_neg_mAb	SCM_neg_VL	0.76	0.58
	N_neg_VH	0.76	0.49	N_phobic_VL	net charges_VH	0.79	0.56