(UroToday.com) The 2024 American Society of Clinical Oncology (ASCO) Annual Meeting will be held in Chicago, Illinois from May 31st.st June 4thNumberA poster session on kidney and bladder cancer was held on November 24, 2024. Dr. Niha Beig presented a machine learning derived model that uses histological features to predict transcriptomic molecular subtypes of advanced renal cell carcinoma (RCC).
Metastatic renal cell carcinoma is a molecularly heterogeneous disease with different levels of angiogenesis, immune infiltration, and PD-L1 expression. Transcriptome analysis from the phase 3 IMmotion151 trial identified seven molecular subtypes that showed distinct outcomes with atezolizumab + bevacizumab and sunitinib treatment.1
In this study, Dr. Beig and colleagues presented histologic correlates of these molecular subtypes identified on hematoxylin and eosin (H&E)-stained whole-slide images of tumors. The study objectives were to develop a machine learning model to derive histologic features of metastatic RCC tumors, identify histologic correlates of RCC molecular subtypes on H&E whole-slide images, and evaluate them as surrogate image-based predictive biomarkers.
In this exploratory analysis, we used whole-slide images to evaluate image-based features in untreated metastatic RCC and their association with molecular subtypes and clinical outcomes. Machine learning models identified 922 H&E-derived human-interpretable histological features in RCC that were associated with cell and tissue morphology, and nuclear shape, in tumor and stroma (including blood vessels, immune cells, and fibroblasts). These machine learning human-interpretable histological features were then extracted from whole-slide images from two metastatic RCC studies, IMmotion151 (n=97, discovery cohort) and IMmotion150 (n=203, validation cohort).2,3
As mentioned above, the seven molecular subtypes were merged into four subgroups to increase computational power.
- Angiogenesis (composed of angiogenic/interstitial and angiogenic)
- Complement/Omega Oxidation
- T Effector
- Proliferative (comprised of proliferative and interstitial proliferative)
The snoRNA subset was excluded from this analysis due to its very low prevalence.
Univariate analysis with false discovery rate (FDR) correction was applied to identify positively correlated human-interpretable features in each of the four subgroups of IMmotion 151 whole slide images, which were then validated in the IMmotion150 molecular subgroups. Representative machine learning human-interpretable features that showed uniquely high abundance in each molecular subgroup in both studies were dichotomized into tertiles of “high” or “low/intermediate” and associated with progression-free survival to fit a Cox proportional hazards model in the IMmotion 151 study.
Each RCC molecular subgroup was associated with a distinct imaging-based phenotype.
- The angiogenesis subgroup had a higher prevalence of 40 image-based features related to the density of endothelial cells and blood vessels within the cancer epithelium.
- T effector subtypes showed a greater abundance of 64 image-based features associated with the presence of immune cells within the stroma.
- The proliferative subgroup had a higher prevalence of 40 imaging-based features related to nuclear morphology.
- No imaging-based features specific to the complement/omega oxidation subgroup were demonstrated.
Representative human interpretable signatures enriched for T effector and proliferation subgroups demonstrated improved PFS benefit with atezolizumab + bevacizumab compared with sunitinib.
Dr Baig concluded:
- They identified unique histologic features of RCC tumors that correlated with previously defined molecular subtypes and were associated with distinct clinical outcomes.
- These results suggest that clinically relevant RCC subtypes can be extracted directly from H&E-stained whole-slide images and may complement gene expression-based patient stratification and selection strategies.
- Further prospective validation of the potential of a biomarker-directed approach to 1st Line RCC treatment is warranted.
Presenter: Dr. Niha Beig, Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA.
Author: Rashid Sayyid, MD, MSc – Clinical Fellow, Society of Urologic Oncology (SUO), University of Toronto, @rksayyid on Twitter, 2024 American Society of Clinical Oncology (ASCO) Annual Meeting, Chicago, IL, May 31st – June 4Number2024
References:
- Motzer RJ, Banchereau R, Hamidi H, et al. Molecular subsets in renal cancer determine checkpoint and antiangiogenic outcomes. Cancer Cell. 2020;38(6): 803-17.e4.
- McDermott DF, Huseni MA, Atkins MB, et al. Clinical activity and molecular correlates of response to atezolizumab alone or in combination with bevacizumab and sunitinib in renal cell carcinoma. National Med2018 Jun;24(6):749-757.
- Rini BI, Powles T, Atkins MB, et al. Atezolizumab plus bevacizumab versus sunitinib in patients with previously untreated metastatic renal cell carcinoma (IMmotion151): a multicenter, open-label, phase 3, randomized controlled trial. Lancet 2019 Jun 15;393(10189):2404-2415.
