Brais, L. | Search bioRxiv

bioRxiv2022

Intertumoral lineage diversity and immunosuppressive transcriptional programs in well-differentiated gastroenteropancreatic neuroendocrine tumors

Neuroendocrine tumors (NETs) are rare cancers that may arise in the gastrointestinal tract and pancreas. The fundamental mechanisms driving gastroenteropancreatic (GEP) NET growth remain incompletely elucidated; however, the heterogeneous clinical behavior of GEP-NETs suggests that both cellular lineage dynamics and tumor microenvironment influence tumor pathophysiology. Here, we investigated the single-cell transcriptomes of tumor and immune cells from patients with gastroenteropancreatic NETs. Malignant GEP-NET cells expressed genes and regulons associated with normal, gastrointestinal endocrine cell differentiation and fate determination stages. While tumor and lymphoid compartments sparsely expressed immunosuppressive targets, infiltrating myeloid cells were enriched for alternative immunotherapy pathways including VSIR, Tim3/Gal9, and SIGLEC10. Finally, analysis of paired primary and metastatic tissue specimens from small intestinal NETs demonstrated transcriptional transformation between the primary tumor and its distant metastasis. Our findings highlight the transcriptomic heterogeneity that distinguishes the cellular landscapes of GEP-NET anatomic subtypes and reveal potential avenues for future precision medicine therapeutics.

cancer biology↗

bioRxiv2021

Pancreatic cancer risk predicted from disease trajectories using deep learning

Pancreatic cancer is an aggressive disease that typically presents late with poor patient outcomes. There is a pronounced medical need for early detection of pancreatic cancer, which can be addressed by identifying high-risk populations. Here we apply artificial intelligence (AI) methods to a dataset of 6 million patient records with 24,000 pancreatic cancer cases in the Danish National Patient Registry (DNPR) and, for comparison, a dataset of three million records with 3,900 pancreatic cancer cases in the United States Department of Veterans Affairs (US-VA) healthcare system. In contrast to existing methods that do not use temporal information, we explicitly train machine learning models on the time sequence of diseases in patient clinical histories and test the ability to predict cancer occurrence in time intervals of 3 to 60 months after risk assessment. For cancer occurrence within 36 months, the performance of the best model (AUROC=0.88, DNPR), trained and tested on disease trajectories, exceeds that of a model without longitudinal information (AUROC=0.85, DNPR). Performance decreases when disease events within a 3 month window before cancer diagnosis are excluded from training (AUROC[3m]=0.83). Independent training and testing on the US-VA dataset reaches comparable performance (AUROC=0.78, AUROC[3m]=0.76). These results raise the state-of-the-art level of performance of cancer risk prediction on real-world data sets and provide support for the design of prediction-surveillance programs based on risk assessment in a large population followed by affordable surveillance of a relatively small number of patients at highest risk. Use of AI on real-world clinical records has the potential to shift focus from treatment of late-stage to early-stage cancer, benefiting patients by improving lifespan and quality of life.

bioinformatics↗