Table 7.

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 and decision tree (DT) models for classifying low/high viscosity. There are 20 mAbs in this study plus 27 mAbs from the literature. The ACC and AUPRC are averaged from 100 randomly generated 4-fold cross-validation sets. The baseline ACC is 0.74 and the baseline AUPRC is 0.26

	One-feature	ACC	AUPRC	Two-features		ACC	AUPRC
	mAbCSP	0.81	0.49	N_phobic_VL	net charges_VL	0.85	0.60
	net charges_VL	0.76	0.39	N_phobic_VL	mAbCSP	0.85	0.58
LR	N_neg_VL	0.77	0.37	net charges_VL	HVI	0.84	0.56
	FvCSP	0.76	0.36	N_phobic_Fv	net charges_VL	0.84	0.56
	N_pos_VL	0.75	0.35	N_phobic_mAb	net charges_VL	0.83	0.55
	mAbCSP	0.81	0.47	N_phobic_VL	net charges_VL	0.83	0.53
	net charges_mAb	0.77	0.37	N_philic_mAb	mAbCSP	0.83	0.51
SVM	net charges_VL	0.76	0.37	net charges_mAb	mAbCSP	0.83	0.50
	N_pos_VL	0.73	0.29	N_neg_VH	net charges_mAb	0.82	0.49
	net charges_VH	0.75	0.28	net charges_VL	net charges_mAb	0.82	0.49
	net charges_mAb	0.78	0.47	N_neg_Fv	net charges_VL	0.85	0.57
	N_phobic_VH	0.77	0.42	net charges_VL	net charges_mAb	0.82	0.53
KNN	net charges_VL	0.78	0.42	net charges_VH	net charges_mAb	0.82	0.53
	mAbCSP	0.76	0.41	N_philic_VL	net charges_VL	0.82	0.53
	SAP_pos_VL	0.73	0.39	mAbCSP	HVI	0.80	0.53
	mAbCSP	0.81	0.47	N_phobic_VL	net charges_VL	0.85	0.57
	SAP_pos_mAb	0.75	0.41	net charges_VL	net charges_mAb	0.84	0.56
DT	net charges_mAb	0.75	0.40	N_philic_VL	net charges_VL	0.84	0.54
	net charges_VL	0.76	0.39	SAP_pos_mAb	FvCSP	0.78	0.48
	net charges_VH	0.76	0.35	SCM_pos_VL	mAbCSP	0.80	0.48