Artificial Intelligence in Oncology - Supporting scientific research
Erasmus Medical Center, Rotterdam
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.
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.