Pathway-Level Prediction and Gene Anchoring in PDAC Identify a TST-Linked Fatty-Acid/Glutathione Axis
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background/Objectives: Metabolic rewiring and an immunosuppressive microenvironment are central to PDAC, yet transcriptome-based measures that are accurate, interpretable, and transferable remain limited. This study aimed to identify which immunometabolic pathways can be predicted consistently from routine tumor transcriptomes and to provide gene-level anchors that clarify the biology. Methods: A predefined panel of 10 metabolic and 5 immune pathways was analyzed in a primary PDAC cohort and validated externally. Pathway activities were summarized by gene-set methods, and immune composition was estimated by digital cytometry. For each pathway, supervised models (Elastic Net, Random Forest, XGBoost) predicted pathway activity from expression with or without microenvironment inputs. Robustness was assessed by permutation tests in single-feature and correlation-aware groupwise modes, and gene–pathway correlations supplied molecular anchors. Results: Metabolic pathways were predicted accurately from expression, and this pattern reproduced in external validation. Immune pathways showed more variable predictability, with improvement when microenvironment composition was included. Robustness was supported by sizable performance losses after permuting top features and by stable ranking under correlation-aware permutation. Among metabolic targets, fatty-acid and sterol biosynthesis and glutathione metabolism showed the highest predictability. TST (thiosulfate sulfurtransferase) emerged as the top shared gene anchor for fatty-acid biosynthesis and glutathione metabolism, linking lipid-anabolic demand to antioxidant capacity in PDAC. Conclusions: A compact, pathway-centered framework was established that pairs predictive modeling with gene-level interpretability and includes external validation. Key metabolic activities can be recovered reliably from routine transcriptomes, whereas immune activities benefit from microenvironment context. The TST-linked fatty-acid/glutathione axis provides a biologically interpretable entry point for mechanistic and translational follow-up.