Prospective inclusion of 1000 patients with pancreatic cancer (early-stage pancreatic cancer accounts for approximately 75% of cases), 1000 patients with benign pancreatic diseases, and 1000 healthy individuals as controls. Peripheral blood samples were collected from newly diagnosed pancreatic cancer patients and healthy individuals. Using techniques such as plasma TCR/BCR-seq, CyTOF, and plasma proteomics, multi-modal individual immune characteristics were obtained and analyzed along with clinical information. An artificial intelligence predictive model was built based on these multi-modal individual immune characteristics to establish an early screening technique for pancreatic cancer. The sensitivity and specificity of this artificial intelligence model for early pancreatic cancer diagnosis were evaluated using an external multicenter sample test set.

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive cancers and currently ranks as the seventh leading cause of cancer-related deaths in China. The nonspecific symptoms of early PDAC are one of the most significant reasons for its low 5-year survival rate. Additionally, PDAC lacks highly sensitive and specific biomarkers for early detection and preventive screening. For the majority of advanced PDAC cases, pathological confirmation often requires tissue biopsies obtained through invasive procedures, and due to the scarcity of tumor cells in these biopsies, pathology may be unclear in up to 20% of cases. On the other hand, there are currently no effective and targeted treatments for PDAC, with surgical resection being the only available option. However, this is only applicable to a small fraction of early-stage PDAC patients, as more than 80% of PDAC patients are diagnosed with distant metastasis at initial diagnosis, where only adjuvant therapy is feasible.

Early diagnosis and detection of PDAC can significantly improve patient prognosis. To date, the only diagnostic biomarker for PDAC is serum Carbohydrate antigen199(CA199) levels, which are neither diagnostic nor specific. High CA199 levels are uncommon in early PDAC but most common in late-stage PDAC. Furthermore, elevated CA199 levels are frequently detected in various benign and malignant diseases, including pancreatitis, cholestasis, and cancer. Additionally, due to screening limitations, imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and endoscopic ultrasound (EUS) are insufficient for early detection of PDAC.

Ideally, the investigators aim to obtain non-invasive, reliable, and repeatable biological markers with clinical potential for early cancer diagnosis. Early studies have identified circulating tumor DNA (ctDNA), circulating tumor cells (CTC), extracellular vesicles (EV), plasma proteomics, and circulating tumor cells (CTCs) as promising real-time and remote tools for this purpose. Compared to other solid tumors, especially lung and breast cancer, some of these circulating biomarkers have entered clinical practice, while blood-derived biomarkers for PDAC diagnosis or monitoring are very limited (excluding CA199), and CA199 is largely underdeveloped compared to other tumors. One example is the use of EpCAM(Epithelial cell adhesion molecule) and cytokeratin for CellSearch system diagnostics, an FDA(Food and Drug Administration)-approved method for diagnosing metastatic breast, colon, and prostate cancers, which was evaluated for PDAC diagnosis, achieving an accuracy rate of 11-78.5%, indicating a wide variation in PDAC detection rates. Other molecular features for diagnosing PDAC, including KRAS mutations in CTCs, miRNA in EVs, and heparan sulfate proteoglycan glypican 1 (GPC1) in extracellular vesicles, unfortunately, exhibit significant differences in sensitivity and predictive performance across different studies, particularly between tumor and CTC status. One study found that in 97% of patients with tumors carrying mutated KRAS, only 18% of CTCs carried the wild-type KRAS allele, even from metastatic tumors. Therefore, due to genetic limitations associated with CTC enrichment and identification, ctDNA isolation, etc., using a single biomarker may only capture partial tumor biological characteristics, leading to low consistency and false negatives. Given the challenges of early diagnosis of PDAC, it is imperative to develop one or more new biological markers for early PDAC diagnosis to capture more possible biological characteristics of primary and metastatic tumors.

The immune system is an extremely important defense system in the human body. Through specific and non-specific biological processes, the immune system can detect various pathogens and harmful substances ranging from viruses to parasites, and differentiate them from healthy cells and tissues in the body under normal circumstances. Therefore, there is a close connection between the immune system and the overall health of an individual. Tumor development remains under constant surveillance by the immune system, and in the early stages of disease, the immune system responds and immune features undergo changes. However, when clinical symptoms appear, it indicates that the immune system is struggling to overcome the presence of harmful substances, and at this point, the disease has already progressed to the middle or late stage. By establishing a large-scale early pancreatic cancer clinical cohort, collecting high-quality multi-modal immune data from individuals, and integrating cutting-edge artificial intelligence models, it is possible to decode individual immune features and develop early diagnostic techniques for pancreatic cancer based on capturing immune status and response signals. This approach can help identify early risk factors for pancreatic cancer by analyzing abnormal immune responses before disease progression, enabling early diagnosis of pancreatic cancer.

Inclusion Criteria:

• Sign the informed consent form;
• Initial diagnosis as patients with pancreatic cancer, patients with benign pancreatic lesions, or healthy controls.

Exclusion Criteria:

• History of other malignancies;
• Presence of organ dysfunction;
• Concurrent immunodeficiency syndrome, active tuberculosis, HIV infection, etc.;
• Allogeneic transplantation requiring immunosuppressive therapy;
• Poor follow-up compliance.

Prospective inclusion of 1000 patients with pancreatic cancer (early-stage pancreatic cancer accounts for approximately 75% of cases), 1000 patients with benign pancreatic diseases, and 1000 healthy individuals as controls. Peripheral blood samples were collected from newly diagnosed pancreatic cancer patients and healthy individuals. Using techniques such as plasma TCR-seq, CyTOF, and plasma proteomics, multi-modal individual immune characteristics were obtained and analyzed along with clinical information. An artificial intelligence predictive model was built based on these multi-modal individual immune characteristics to establish an early screening technique for pancreatic cancer.