Fig. 4: Diagnosis of PDAC with clinical serum samples using ELISA, LFIA, and SELFI.

a Schematic illustration of the statistical data analysis process. b Average ROC curve for ELISA, LFIA, and SELFI using samples from early-stage PDAC patients and healthy controls (area under the curve (AUC): ELISA = 0.854, LFIA = 0.550 and SELFI = 0.862). c Average ROC curve for ELISA, LFIA, and SELFI using samples from late-stage PDAC patients and healthy controls (AUC: ELISA = 0.926, LFIA = 0.692 and SELFI = 0.925). d Average ROC curve for ELISA, LFIA, and SELFI using total samples (AUC: ELISA = 0.896, LFIA = 0.634 and SELFI = 0.899). e Violin plot of AUC values for ELISA, LFIA, and SELFI using samples from early-stage PDAC patients and healthy controls (P < 0.0001, two-sided Student’s t test). f Violin plot of AUC values for ELISA, LFIA, and SELFI using samples from late-stage PDAC patients and healthy controls. Two-sided Student’s t test shows a significant difference between LFIA and SELFI (P < 0.0001), but not between ELISA and SELFI (ns; P = 0.3363). g Violin plot of AUC values for ELISA, LFIA, and SELFI using total samples. Statistical significance was assessed using a two-sided Student’s t test (ELISA vs. SELFI, P = 0.0142; LFIA vs. SELFI, P < 0.0001). For b–g, data from ELISA, LFIA, and SELFI are depicted in orange, red, and blue, respectively. The ROC curves and AUC values were obtained using a homemade R program code with the pROC library. For b–d, the average ROC curves were obtained by averaging 100 ROC curves using linear interpolation, with the black dotted line representing a random classifier. For e–g, the median value is depicted as a solid line, and the interquartile range boundaries are depicted as dotted lines in each violin plot.