Deep learning-based CD8⁺ T cell model for predicting prognosis and‌ targeted immunotherapy benefit in ccRCC

Area of Science:

• Oncology
• Pathology
• Artificial Intelligence
• Immunology

Background:

• Deep learning shows potential for enhancing clear cell renal cell carcinoma (ccRCC) diagnosis and treatment.
• Research on the tumor microenvironment (TME) using pathological images and AI in ccRCC is limited.
• Accurate assessment of CD8+ T cell infiltration is crucial for ccRCC prognosis and immunotherapy response.

Purpose of the Study:

• To develop and validate a deep learning-based CD8+ T cell inflammation signature (DL-CD8T) for ccRCC using H&E-stained whole slide images.
• To assess the association of DL-CD8T with TME characteristics, clinical outcomes, and treatment response in ccRCC.
• To explore the potential of DL-CD8T in improving risk stratification and guiding personalized therapy for ccRCC.

Main Methods:

• A multi-center cohort study utilizing a clustering-constrained attention multiple-instance learning method.
• Development of the DL-CD8T model on a training cohort using H&E-stained whole slide images.
• Independent validation of the DL-CD8T model in the CPTAC cohort.

Main Results:

• The DL-CD8T model achieved an AUC of 0.781 in the training cohort and 0.741 in the independent CPTAC cohort for distinguishing high CD8+ T cell inflammation.
• DL-CD8T positivity correlated significantly with high CD8+ T cell infiltration, elevated PD-1, PD-2, PD-L1, CTLA-4, TIM-3 expression, and higher tumor mutation burden.
• DL-CD8T identified ccRCC patients with high survival risk and showed potential in predicting response to combined targeted and immunotherapy (HR=0.27, p=0.029).

Conclusions:

• The developed DL-CD8T model effectively identifies CD8+ T cell inflammation from H&E images in ccRCC.
• DL-CD8T serves as a valuable biomarker for risk stratification and predicting immunotherapy response in ccRCC.
• DL-CD8T holds promise for personalized targeted therapy in ccRCC, pending prospective validation.