1143O - Refining molecular classification of neuroendocrine tumors from diverse origins using multi-omic integration models: Unveiling novel neuroendocrine subtypes and their clinical implications

Background

Neuroendocrine tumors (NETs) represent a rare spectrum of tumors with a wide anatomical distribution, sharing common clinical and molecular traits. While NETs exhibit few recurrent mutations, epigenetics play a significant role. However, a comprehensive molecular understanding across different primary sites is still pending. Our study aimed to develop a molecular classification of NETs and identify shared vulnerabilities.

Methods

Transcriptomic and methylomic profiles from 194 paraffin-embedded tumor samples of patients with NETs from different primary sites within the GEP tract and lungs were analyzed. Using Multi-Omics Factor Analysis (MOFA), an 8-factor model integrating both omics datasets was constructed to apply clustering algorithms. Differences in discrete variables among subtypes were evaluated by Fisher's test, and Kaplan-Meier and Cox regression models were used to assess overall survival. Immune cell populations were characterized via deconvolution and differential expressed genes and Gene Set Enrichment Analysis were conducted.

Results

The MOFA model generated identified three neuroendocrine subtypes (NS) significantly linked to patient prognosis. NS1 showed good prognosis, characterized by an enrichment in diverse immune populations and increased expression of metabolism and nutrient digestion genes. NS3 was associated with more aggressive clinical features (G2-3, stage IV at diagnosis, higher 5-HIAA levels) and had the worst prognosis. NS3 exhibited elevated levels of methylation and copy number variations, and higher expression of genes associated with neuroendocrine function and with cell proliferation such as PTPRN, BEX1, and NOVA1. NETs with intermediate prognosis, NS2, presented a mixed molecular profile.

Conclusions

This molecular classification unveils common sources of biological variability among different primary tumor origins and classify samples into three prognostic NS associated with distinct molecular features. Additionally, this study may facilitate the identification of universal prognostic biomarkers and may enable the development of common therapeutic strategies for NETs.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

GETNE, Comunidad de Madrid, AECCC, ISCIII.

Disclosure

All authors have declared no conflicts of interest.