Table 5.

Classification results of CA19-9 only, variants (with deep learning), and CA19-9 and variants together (with deep learning). Variants together with CA19-9 achieved very high performance, especially for rPDAC vs. CP.

Task	Variable		Performance on the Test Set
	Type	No of Features	AUC	Precision	Recall	F1	Accuracy
Cancer vs. CP	CA19-9	1	0.84	0.89	0.72	0.79	0.75
	Variants	70	0.83	0.76	0.83	0.79	0.71
	CA19-9 & Variants	71	0.96	0.89	0.97	0.93	0.90
rPDAC vs. CP	CA19-9	1	0.85	0.78	0.72	0.75	0.77
	Variants	76	0.89	0.88	0.88	0.88	0.88
	CA19-9 & Variants	77	0.96	0.93	0.88	0.90	0.91
nrPC vs. CP	CA19-9	1	0.84	0.80	0.72	0.76	0.76
	Variants	67	0.76	0.68	0.89	0.77	0.72
	CA19-9 & Variants	68	0.92	0.75	0.95	0.84	0.81