Related Experiment Video
Updated: Jun 25, 2026
06:51
A Preclinical Murine Model of Hepatic Metastases
Published on: September 27, 2014
28.7K
Developing a Predictive Model for Metastatic Potential in Pancreatic Neuroendocrine Tumor
Jacques A Greenberg1, Yajas Shah2, Nikolay A Ivanov1
1Department of Surgery, Weill Cornell Medicine, New York, NY 10065, USA.
The Journal of Clinical Endocrinology and Metabolism
|May 31, 2024
Summary
Researchers developed an 8-gene panel using machine learning to predict pancreatic neuroendocrine tumor (PNET) metastasis. This transcriptomic signature accurately classifies tumor metastatic potential, aiding clinical management decisions.
Area of Science:
- Oncology
- Genomics
- Machine Learning
Background:
- Pancreatic neuroendocrine tumors (PNETs) display variable clinical behavior, ranging from localized disease to aggressive metastasis.
- A comprehensive transcriptomic profile to differentiate PNET phenotypes is currently lacking.
Purpose of the Study:
- To develop machine learning-based predictive models for PNET metastatic potential.
- To identify a transcriptomic signature capable of distinguishing between localized and metastatic PNETs.
Main Methods:
- RNA-sequencing data from 95 primary PNETs were analyzed.
- Machine learning models were trained to predict metastatic status and validated on an independent cohort using NanoString nCounter®.
Main Results:
- An 8-gene panel (AURKA, CDCA8, CPB2, MYT1L, NDC80, PAPPA2, SFMBT1, ZPLD1) was identified as sufficient for classifying metastatic PNETs.
- The models achieved high sensitivity (87.5-93.8%) and specificity (78.1-96.9%) in predicting metastatic status.
- Validation on an independent cohort demonstrated the model's predictive accuracy with a median AUC of 0.886.
Conclusions:
- An 8-gene panel accurately predicts the metastatic phenotype in PNETs.
- This panel is detectable using the clinically available NanoString nCounter® system.
- Prospective studies are warranted to evaluate its utility in guiding patient management.
