Multi-protein panel in pancreatic cyst fluid for improved risk stratification for pancreatic cancer

Pancreatic cystic lesions (PCLs) are the sole radiologically recognisable and highly heterogeneous precursors of pancreatic cancer (PC). The malignant potential of PCLs is inferred from their types, as determined by empirical clinical practice guidelines; however, accurate risk stratification of patients preoperatively presents an unmet clinical need. We performed deep proteomic profiling of pancreatic cyst fluid (PCyF) and identified a first-of-its-kind multi-protein (n=89) panel termed "ASSIGN1" - Early diagnosis and detection of pAncreatic cySt malignancy SIGNature. ASSIGN1 was used for the development and validation of a support vector machine-based model for predicting malignant potential (based on malignancy risk score, zero to one) of individual PCLs using discovery/training and validation/test cohorts. The diagnostic accuracy of the model was evaluated based on histopathology of resected cysts using sensitivity, specificity and area under the receiver-operating-characteristic (AUROC) curve measures and compared to Fukuoka guidelines-based preoperative assessment. ASSIGN1-based malignancy risk score was a cyst type-independent and accurate (sensitivity=1.00, specificity=1.00 and AUROC=1.00) predictor of (i) pancreatic carcinoma and (ii) malignant potential of PCLs, which outperformed international consensus Fukuoka guidelines-based preoperative assessment (sensitivity=1.00; specificity=0.38; AUROC=0.71). Our findings demonstrated that ASSIGN1 holds promise to replace current preoperative laboratory tests, complement existing standard-of-care practices and improve preoperative diagnosis of PCLs and early detection of PC.