Related Experiment Video
Updated: Oct 19, 2025
07:41
A Robust Discovery Platform for the Identification of Novel Mediators of Melanoma Metastasis
Published on: March 8, 2022
2.6K
Integrating 31-Gene Expression Profiling With Clinicopathologic Features to Optimize Cutaneous Melanoma Sentinel
Eric D Whitman1, Vadim P Koshenkov2, Brian R Gastman3
1Carol G. Simon Cancer at Morristown Medical Center, Atlantic Health System, Morristown, NJ.
JCO Precision Oncology
|September 27, 2021
Summary
An artificial intelligence algorithm integrating clinicopathologic features with the 31-gene expression profile (31-GEP) score improves sentinel lymph node biopsy (SLNB) risk prediction for melanoma patients. This tool helps identify patients who may not need SLNB and those more likely to benefit from it.
Area of Science:
- Oncology
- Genomics
- Artificial Intelligence
Background:
- National guidelines recommend sentinel lymph node biopsy (SLNB) for melanoma patients with >10% risk of sentinel lymph node (SLN) positivity.
- SLNB is not recommended for T1a melanoma with <5% SLN positivity risk.
- SLNB decision-making is uncertain for higher-risk T1 melanomas with a 5%-10% SLN positivity risk.
Purpose of the Study:
- To develop and validate an artificial intelligence (AI) algorithm integrating clinicopathologic features with the 31-gene expression profile (31-GEP) score for precise SLN positivity risk prediction.
- To improve clinical decision-making regarding SLNB for melanoma patients, particularly those with intermediate risk.
Main Methods:
- An integrated 31-GEP (i31-GEP) neural network algorithm was developed using a cohort of 1,398 patients.
- The i31-GEP algorithm was validated using an independent cohort of 1,674 patients.
- Clinicopathologic features were integrated with the continuous 31-GEP score using AI.
Main Results:
- The continuous 31-GEP score was the strongest predictor of SLN positivity (G² = 91.3, P < .001).
- The i31-GEP demonstrated high concordance between predicted and observed SLN positivity rates (slope = 0.999).
- The i31-GEP increased the proportion of patients predicted with <5% SLN positivity from 8.5% to 27.7% (NPV 98%), and reclassified 63% of intermediate-risk T1 melanoma cases.
Conclusions:
- The i31-GEP algorithm provides a more precise, personalized, and clinically actionable SLN-positive likelihood estimate.
- This AI tool has the potential to reduce unnecessary SLNBs by identifying patients below the 5% risk threshold.
- The i31-GEP may improve SLNB yield by better identifying patients likely to have a positive SLN.
