GAN-enhanced Machine Learning and Metabolic Modeling Identify Reprogramming in Pancreatic Cancer

Authors

• Tahereh Razmpour, University of Nebraska-Lincoln
• Masoud Tabibian, University of Nebraska-Lincoln
• Arman Roohi, University of Illinois at Chicago
• Rajib Saha, University of Illinois at ChicagoFollow

ORCID IDs

Roohi https://orcid.org/0000-0002-0900-8768

Saha https://orcid.org/0000-0002-2974-0243

Date of this Version

2026

Document Type

Article

Citation

PLoS Computational Biology (2026) 22(1): e1013862

doi: 10.1371/journal.pcbi.1013862

Comments

Open access

License: CC BY 4.0

Abstract

Pancreatic ductal adenocarcinoma is one of the deadliest forms of cancer, presenting significant clinical challenges due to poor prognosis and limited treatment options. Understanding the metabolic reprogramming that drives this disease is crucial for identifying new therapeutic targets and improving patient outcomes. We developed a novel computational framework integrating genome-scale metabolic modeling with machine learning to identify metabolic signatures and therapeutic vulnerabilities in pancreatic cancer. To address the inherent class imbalance in cancer datasets, we generated synthetic healthy samples using a Wasserstein Generative Adversarial Network with Gradient Penalty, implementing a rigorous three-step biological filtration process to ensure their validity. This approach enabled the creation of a balanced dataset for robust comparison of healthy versus cancerous metabolic states. Our machine learning classifier achieved 94.83% accuracy in distinguishing between these states, demonstrating the effectiveness of our integrated approach. Systems-level analysis revealed three key dysregulated pathways: heparan sulfate degradation, O-glycan metabolism, and heme degradation. We identified impaired lysosomal degradation of heparan sulfate proteoglycans as a potential contributor to disease pathogenesis, providing a mechanistic explanation for the previously observed association between lysosomal storage disorders and pancreatic cancer. Additionally, nervonic acid transport emerged as the most discriminative reaction between healthy and cancerous states, with gene-level analysis highlighting fatty acid binding proteins, fatty acid transporters, and acyl-CoA synthetases as key molecular drivers of metabolic reprogramming. Our multi-level approach connected genetic drivers to functional metabolic consequences, revealing coordinated upregulation of fatty acid transport and activation processes. These findings enhance our understanding of pancreatic cancer metabolism and present potential therapeutic targets, demonstrating the value of integrated computational approaches in cancer research.