Casper van Eijck

Erasmus Medical Center, Rotterdam

Clinical Decision Support for Pancreatic Cancer

Pancreatic adenocarcinoma (PDAC) is the fourth cause of cancer-related mortality world-wide with a five-year survival rate of less than 5%. Despite multiple large-scale genetic sequencing studies, identification of predictors of treatment response and patient survival remains challenging. The advance of artificial intelligence enables us to decipher the relationship between various levels of genetic measurements and the patient outcomes, and gain a more complete picture of the disease to develop better prognostic strategies to stratify patients. We aim to develop artificial intelligence frameworks that can be applied to support clinical decision-making in prospective clinical trials and to improve patient stratification by targeting those patients that will benefit from FOLFIRINOX treatment. This will improve treatment efficacy and reduce toxicity by withholding ineffective treatments for selected patients.

Brief summary of progress / final results august 2024

The high heterogeneity of pancreatic cancer has posed big challenges for prognosis prediction and treatment management. In this project, we developed artificial intelligence frameworks that can support PDAC clinical decision-making and improve PDAC patient stratification by targeting patients who will benefit from heavy treatments such as surgery or FOLFIRINOX. More specifically, we developed a deep learning-based framework to predict the survival of PDAC patients. We identified corresponding multi-omics biomarkers related to DNA damage repair and immune response processes. The biomarkers were validated in independent datasets. To our knowledge, our study is the first attempt to use deep learning for PDAC data integration.

Further, to bring computational modeling using artificial intelligence closer to clinical practice, we created a collection of machine learning workflows on the Galaxy platform for clinicians and biologists to build and deploy predictive classifiers. Being freely available and accessible, it automates data analysis and enables clinicians and researchers to perform various predictive tasks. Another artificial intelligence technique, transfer learning, is also explored to utilize the knowledge shared among broader cancer types for survival and drug response prediction. This alleviates the shortage of available data given this rare cancer.

Moreover, we contributed to biomarker discovery and understanding of the biological mechanisms, including defining marker genes, identifying circulating prognostic biomarkers, studying the PDAC tumor immune microenvironment, and exploring the FOLFIRINOX effect on tumor morphology and spatial transcriptomics. We believe the data, knowledge, and tools generated in this project provide valuable insight and utilities for pancreatic cancer patient outcome prediction, thus helping clinicians in treatment decision-making to gauge the benefits versus the risks better. We thank the Hanarth Foundation for their financial support.

Brief summary of progress / results November 2022

A major challenge for treating pancreatic ductal adenocarcinoma (PDAC) patients is the unpredictability of their prognoses due to high heterogeneity. We present Multi-Omics Deep Learning for Prognosis-correlated subtyping (MODEL-P) to identify PDAC subtypes and to predict prognoses of new patients. MODEL-P was trained on a deep learning model which integrates multi-omics of 146 PDAC patients together with their survival outcome. Using MODEL-P, we identified two PDAC subtypes with distinct survival outcomes (median survival 10.1 and 22.7 months respectively), which correspond to DNA damage repair and inhibited immune response. We rigorously validated MODEL-P by stratifying patients in five independent datasets into these two survival groups and achieved significant survival difference which is superior to current practice and other subtyping schemas. We believe the subtype-specific signatures would facilitate PDAC pathogenesis discovery, and MODEL-P can provide clinicians the prognoses information in the treatment decision-making to better gauge the benefits versus the risks. This work has been published in iScience.

Furthermore, we are experimenting to apply modern AI strategies such as transfer learning and multi-task learning to tackle the problem of not having a big dataset to build a robust prediction model. Meanwhile, we also contribute to the biological understanding of the treatment of PDAC and the statistical soundness of clinical study designs.

Brief summary of progress / results October 2021

Dutch only; please click here.