March 2024 | George M. Pikler, M.D., Ph.D., FACP, Lead Oncology Advocate N1X10

AI Predicts Pancreatic Cancer Survival

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive malignancies. It is estimated that 55,550 deaths from this malignancy will occur in the United States this year and is expected to become the second-leading cause of cancer-related deaths nationally by 2030. While only 30–40% of patients with PDAC present with localized disease and undergo potentially curative surgical resection after diagnosis or following neoadjuvant chemotherapy, most develop recurrence and succumb to their disease.

Despite advancements in molecular testing, serum carbohydrate antigen 19-9 (CA 19-9), first discovered in 1979, is presently the only US Food and Drug Administration (FDA)-approved biomarker widely employed for diagnostic management and preoperative prognostication of PDAC. CA 19-9 however, has limitations, with a high false-positive rate due to other pathologic conditions and can result in false negatives in about 10% of the population.

In a recent publication, researchers describe an artificial intelligence (AI) platform known as the Molecular Twin consisting of advanced machine-learning models (ML) which were used to analyze a dataset of 6,363 clinical and multi-omic molecular features across 74 patients (clinical stage I and II) with resected pancreatic ductal adenocarcinoma to accurately predict disease survival (DS). They showed that a full multi-omic model predicts DS with the highest accuracy and that plasma protein is the top single-omic predictor of DS.

Molecular profiling data were collected from both tumor and host samples and included computational pathology features. Multiple ML models were developed and applied to this dataset to test the hypothesis that this approach can provide biomarker panels that accurately predict DS after surgery in patients with resectable PDAC. Utilizing external samples/data from The Cancer Genome Atlas (TCGA), Johns Hopkins University (JHU) and Massachusetts General Hospital (MGH), they independently validated the power of their full and parsimonious ML models to predict DS. Through this analysis, they also discovered that among all analytes available in the preoperative setting, plasma protein is the most critical biomarker with substantial predictive power for survival and superior to CA 19-9.

While their research centered on PDAC, this approach is tumor-type agnostic, allowing it to potentially impact clinical care and scientific discovery across all cancers.

Nature Cancer. 2024 (5); 299-414

Erica Cross, PA


Erica is a board certified Physician Assistant. She obtained her Master’s degree in Physician Assistant studies from Our Lady of the Lake College in Baton Rouge, LA. She began practicing in 2011 and has worked clinically in Orthopedics and Dermatology. The majority of her career has been spent in a Dermatology practice where she assisted in Mohs surgery, treating various types of skin cancer. She also teaches in the medical simulation department at the University of South Alabama and enjoys every aspect of medical education.