Machine Learning Helps Predict Which Patients Respond to Immunotherapy

A machine learning algorithm helped to predict which patients with melanoma respond better to immunotherapy and who would not respond, according to a study by researchers at the Johns Hopkins Kimmel Cancer Center and its Bloomberg Kimmel Institute for Cancer Immunotherapy.

The DeepTCR system showed efficacy as a predictive clinical tool, but also provided information on the biological mechanisms that underlie how patients respond to immunotherapy.

“DeepTCR’s predictive power is exciting, but what I found more fascinating is that we were able to view what the model learned about the immune system’s response to immunotherapy,” said John-William Sidhom, MD, PhD, first author of the study, in a press release. “We can now exploit that information to develop more robust models, and possibly better treatment approaches, for many diseases, even those outside of oncology.”

DeepTCR uses deep learning to recognize patterns in large volumes of data from amino acid sequences of proteins called T cell receptors (TCRs). These sit on the exterior of the immune system’s T cells while waiting to engage a protein from an enemy, such as bacteria or viruses.

Current immunotherapy drugs, or checkpoint inhibitors, involve proteins that confuse this capacity in tumors, causing the T cells to respond to cancer; however, these drugs have been found to help a limited number of patients, according to the investigators.

The current study used materials collected during the CheckMate 038 clinical trial, which evaluated the efficacy of 1 immunotherapy drug (nivolumab) compared to a combination of 2 (nivolumab and ipilimumab) for 43 patients with inoperable melanoma.

Biopsies of the tumors were taken prior to and during treatment. In the study, no significant differences were observed in patients administered the single drug versus the 2-drug combination. Some patients in both groups responded, whereas others did not.

The investigators used high-tech genetic sequencing to evaluate the TCR collection surrounding each tumor by determining the type and number of TCRs in each biopsy. These data were entered into the DeepTCR program along with which data sets belonged to responders versus nonresponders, and then the algorithm looked for patterns.

The researchers evaluated differences prior to treatment between the TCR collection of immunotherapy in responders and nonresponders. The differences identified were as predictive of patient response as known biomarkers, according to the study. The investigators said these results need to be confirmed in a larger patient population before the algorithm can be used to guide therapy.

"Precision immunotherapy based on the immune microenvironment in the tumor is critical to guide the optimal choice of treatment options for each patient,” said Drew Pardoll, MD, PhD, professor of oncology and director of the Bloomberg Kimmel Institute for Cancer Immunotherapy, in a press release. “These DeepTCR findings define a new dimension for predicting a tumor’s response to immune checkpoint blockade by applying a novel artificial intelligence strategy to deconvolute the vast array of receptors expressed by tumor-infiltrating T cells, the key immune components responsible for direct killing of tumor cells.”

The researchers also sought to determine the differences between responders and nonresponders by using data from another study that linked specific TCRs to the enemy proteins that activated them. They found that those who responded to immunotherapy had a higher number of virus-specific T cells in their tumors, and nonresponders had more tumor-specific T cells.

The team learned that the nonresponders had higher turnover of T cells.

“Both responders and nonresponders had about the same number of tumor-specific T cells before and during therapy,” Sidhom said in a press release. “The identity of those T cells remained the same in the responders, but in the nonresponders, there was a different variety of T cells before and during therapy. Our hypothesis is that nonresponders had a high number of ineffective tumor-specific T cells from the start. When the immunotherapy began, their immune systems sent in a new batch of T cells, trying to find an effective one, but the dysfunction remained. On the other hand, the responders had effective T cells from the onset, but their anti-tumor activity was blocked by the tumor. When the immunotherapy began, it released the blockade and allowed them to do their job.”

REFERENCE

Machine Learning Can Help Predict Patient Response to Cancer Immunotherapy. Johns Hopkins Medicine. November 16, 2022. Accessed November 17, 2022. https://www.hopkinsmedicine.org/news/newsroom/news-releases/machine-learning-can-help-predict-patient-response-to-cancer-immunotherapy