Related Experiment Video
Updated: Mar 1, 2026
08:50
Predictive Immune Modeling of Solid Tumors
Published on: February 25, 2020
7.6K
Annotation-free prediction of immunotherapy response in melanoma using single-cell transcriptomic data
Da Eun Oh1,2, Gaeun Kee3,4, Ji-Hye Oh1
1Bioinformatics Core Laboratory, Convergence Medicine Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea.
Plos One
|February 27, 2026
Summary
An artificial intelligence model predicts immune checkpoint inhibitor response in melanoma using single-cell RNA sequencing data. This approach identifies novel biomarkers like CCR7, improving patient stratification for personalized immunotherapy.
Area of Science:
- Oncology
- Immunology
- Bioinformatics
Background:
- Immune checkpoint inhibitors (ICIs) have revolutionized advanced melanoma treatment, but response prediction remains challenging due to tumor heterogeneity.
- Existing biomarkers like PD-L1 and tumor mutational burden have limited predictive accuracy for melanoma.
- There is a critical need for novel biomarkers to identify patients likely to benefit from ICIs.
Purpose of the Study:
- To develop and validate an artificial intelligence (AI)-based model for predicting ICI responsiveness in melanoma using single-cell RNA sequencing (scRNA-seq) data.
- To identify novel predictive biomarkers for ICI response without requiring cell type annotation.
- To explore the functional roles and clinical significance of identified biomarkers in melanoma.
Main Methods:
- An AI model was trained on scRNA-seq data from 16,290 tumor-infiltrating cells in melanoma.
- Multiple AI algorithms were evaluated, including Extreme Gradient Boosting, Random Forest, and Convolutional Neural Networks.
- The best-performing model was validated using three independent bulk RNA-seq datasets.
Main Results:
- The AI model achieved an area under the curve of 0.87 in predicting ICI response.
- The model identified 29 predictive biomarkers, notably CCR7 and MTRNR2L2.
- CCR7 was associated with favorable ICI response and improved survival, primarily expressed in B cells and memory T cells of responders.
- MTRNR2L2 was linked to non-response and poorer outcomes, elevated in exhausted T cells of non-responders.
- CCR7 expression was significantly associated with the immune molecular subtype of melanoma.
Conclusions:
- AI-driven predictive modeling of unannotated scRNA-seq data is a robust approach for identifying clinically relevant biomarkers in melanoma.
- The identified biomarkers, particularly CCR7, offer potential for improved patient stratification and personalized immunotherapy strategies.
- This approach facilitates a deeper understanding of the immune landscape in responders versus non-responders to ICI therapy.
