Researchers Develop AI Algorithm to Predict Immunotherapy Response

In the near future, this could make it possible for physicians to use imaging to identify biological phenomena in a tumor located anywhere in the body without performing a biopsy.

Currently, there are no markers, which can accurately identify patients who will respond to anti-PD-1/PD-L1 immunotherapy in a situation where only 15 to 30% of patients do respond to such treatment. The more immunologically richer the tumor environment (presence of lymphocytes), the higher is the chances of immunotherapy being effective. Hence, the researchers tried to characterize this environment using imaging and correlate this with the patients’ clinical response. In their study, the radiomic signature was captured, developed and validated genomically, histologically and clinically in 500 patients with solid tumors (all sites) from four independent cohorts.

The researchers first used a machine learning-based approach to teach the algorithm how to use relevant information extracted from CT scans of patients participating in an earlier study, which also held tumor genome data. Thus, based solely on images, the algorithm learned to predict what the genome might have revealed about the tumor immune infiltrate, in particular with respect to the presence of cytotoxic T-lymphocytes (CD8) in the tumor, thus establishing a radiomic signature.

The researchers tested and validated this signature in other cohorts, including that of TCGA (The Cancer Genome Atlas), thus demonstrating that imaging could predict a biological phenomenon, providing an estimation of the degree of immune infiltration of a tumor. Further, in order to test the signature’s applicability in a real situation and correlate it to the efficacy of immunotherapy, it was evaluated using CT scans performed before the start of treatment in patients participating in five phase I trials of anti-PD-1/PD-L1 immunotherapy. The researchers found that the patients in whom immunotherapy was effective at three and six months had higher radiomic scores as did those with better overall survival.

In their next clinical study, the researchers will assess the signature both retrospectively and prospectively, using a larger number of patients and stratifying them based on cancer type in order to refine the signature. They will also use more sophisticated automatic learning and AI algorithms to predict patient response to immunotherapy, while integrating data from imaging, molecular biology and tissue analysis. The researchers aim to identify those patients who are most likely to respond to treatment, thereby improving the efficacy/cost ratio of treatment.