Predicting Tumor Microenvironment and Immunotherapy Response in Lung Cancer
Author Information
Author(s): Sushant Patkar, Alex Chen, Alina Basnet, Amber Bixby, Rahul Rajendran, Rachel Chernet, Susan Faso, Prashanth Ashok Kumar, Devashish Desai, Ola El-Zammar, Christopher Curtiss, Saverio J. Carello, Michel R. Nasr, Peter Choyke, Stephanie Harmon, Baris Turkbey, Tamara Jamaspishvili
Primary Institution: National Cancer Institute, National Institutes of Health, Bethesda, MD, USA
Hypothesis
Can a deep learning approach accurately predict tumor microenvironment composition and immunotherapy response from histopathology images in non-small cell lung cancer?
Conclusion
The HistoTME model effectively predicts tumor microenvironment signatures and improves prognostication for lung cancer patients receiving immunotherapy.
Supporting Evidence
- HistoTME accurately predicts the expression of 30 distinct cell type-specific molecular signatures.
- The model achieved an average Pearson correlation of 0.5 with ground truth on independent tumor cohorts.
- HistoTME-predicted microenvironment signatures improve prognostication of lung cancer patients receiving immunotherapy.
- The model was validated on a cohort of 652 patients treated with immune checkpoint inhibitors.
- Statistical analysis showed significant differences in overall survival based on TME status.
- Interactions among TME signatures were found to improve prediction of immunotherapy response.
- HistoTME can identify more patients responding to first-line immune checkpoint inhibitor therapy.
- The model provides a cost-effective approach to assess tumor microenvironment using routine pathology slides.
Takeaway
Researchers created a computer program that looks at images of lung cancer to help doctors understand how the tumor environment affects treatment success.
Methodology
The study used a weakly supervised deep learning model trained on histopathology images and transcriptomics data to predict tumor microenvironment signatures.
Potential Biases
Potential bias due to retrospective analysis and lack of external validation datasets.
Limitations
The model's predictions are based on bulk transcriptomics data, which may not capture spatial distribution of cell types.
Participant Demographics
Patients with non-small cell lung cancer, including those treated with immunotherapy.
Statistical Information
P-Value
0.0012
Confidence Interval
[95% CI: 0.61-0.88]
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website