Embracing the Complexity of Somatic Alterations in Kidney Cancer Care
Machine learning-based approaches may play a role in further understanding of how somatic alterations influence responses or resistance to therapy.
In a conversation with CancerNetwork® during a visit to Yale Cancer Center in New Haven, Connecticut, David A. Braun, MD, PhD, detailed his group’s work in understanding how somatic alterations may impact responses or resistance to treatment among patients with kidney cancer.
Braun, assistant professor at Yale School of Medicine and principal investigator in the Center of Molecular and Cellular Oncology within the Yale Cancer Center, stated that it is crucial to embrace the complexity of biomarkers in kidney cancer, as individual genetic alterations alone would not determine whether a tumor responds to therapy. As part of dissecting each potential predictor of response or resistance to treatment, Braun described how approaches such as machine learning-based tools may play a role in advancing the field’s understanding of these factors.
Transcript:
Our research group and many others have investigated this question of how somatic alterations might impact responses or resistance to therapy for many years. The hope, initially, was a simple answer, that we would find something like the EGFR mutation in lung cancer; someone has this mutation, [so] they will respond to a drug, and if they do not have this mutation, they will not. What we have realized over the years is that it’s much more complicated than that. The kidney cancer genetic landscape and the immune microenvironment are both important factors, and that’s not necessarily going to be one thing—one genetic alteration—that’s going to make the difference between whether a tumor responds or doesn’t respond to a treatment.
What we start to move towards is embracing that complexity, understanding that it’s the genetic landscape of the cancer, the immune cells that are there, and other factors that altogether work to determine whether a cancer will effectively respond or not. [Many] of our efforts now have been dissecting each of those individual parts and then using a lot of approaches, including machine learning-based approaches, to put it all together.
