AI-Driven Prediction of Kidney Cancer Treatment Response Shows Promise

Researchers at UT Southwestern Medical Center have developed an advanced artificial intelligence (AI) model capable of predicting which patients with kidney cancer will respond favorably to anti-angiogenic therapy. This class of treatment, which targets the formation of new blood vessels in tumors, is only effective in some cases, and currently, there are no definitive biomarkers to identify those most likely to benefit. The study, published in Nature Communications, demonstrates how AI can analyze histopathological slides — images of tumor tissue sections stained for microscopic examination — to assess tumor vascularity, providing insights akin to genetic tests but with greater accessibility and lower cost.

The team trained their AI system using data sets that linked kidney cancer tissue images with the Angioscore, a gene expression-based biomarker used to predict therapy response. Remarkably, the AI model's visual outputs closely correlated with the Angioscore, indicating its ability to interpret tumor characteristics visible under a microscope.

In testing with over 200 patient samples, the AI approach predicted therapy response nearly as accurately as the Angioscore—correctly identifying responders nearly 73-75% of the time. This method offers a transparent and interpretable visualization of tumor blood vessel density, making it easier for clinicians to understand and trust the predictions.

This breakthrough holds significant potential for personalizing kidney cancer treatment, reducing unnecessary side effects, and improving outcomes. Additionally, the researchers aim to develop similar AI tools to forecast responses to immunotherapy, broadening the scope of precision medicine in oncology.

According to Dr. Satwik Rajaram, the development addresses a critical gap in clinical practice, where readily available pathology slides could be used to guide treatment choices more effectively. Co-lead Dr. Payal Kapur emphasizes that this technology could revolutionize how treatment decisions are made, especially given the heterogeneity of kidney tumors and the limitations of current biomarkers.

Every year, approximately 435,000 new cases of clear cell renal cell carcinoma are diagnosed worldwide. Anti-angiogenic drugs are a common treatment, but less than half of patients derive benefit, exposing many to unnecessary toxicity and financial costs. The new AI tool aims to fill this gap by enabling more accurate, accessible, and cost-effective prediction of therapy response.

The researchers plan to extend their AI platform to predict responses to other therapies, including immunotherapy, paving the way for more personalized and effective cancer care.