New Computational Tool Predicts Immunotherapy Success for Metastatic Breast Cancer Patients

Researchers at Johns Hopkins have unveiled a computational tool designed to predict the success of immunotherapy in patients with metastatic triple-negative breast cancer, marking a significant advance in personalized cancer treatment.

In a breakthrough poised to revolutionize personalized cancer treatment, researchers from the Johns Hopkins Kimmel Cancer Center and the Johns Hopkins University School of Medicine have developed an innovative computational method to identify which patients with metastatic triple-negative breast cancer are likely to benefit from immunotherapy. This pioneering work was recently published in the Proceedings of the National Academy of Sciences.

Immunotherapy harnesses the body’s immune system to combat cancer cells but is effective only for a subset of patients.

“It’s really important that we identify those patients for whom it will work, because the toxicity of these treatments is high,” lead author Theinmozhi Arulraj, a postdoctoral fellow at Johns Hopkins, said in a news release.

The challenge of accurately predicting patient response to immunotherapy has led scientists to explore predictive biomarkers — specific cells or molecules within tumors that signal potential treatment outcomes.

“Unfortunately, existing predictive biomarkers have limited accuracy in identifying patients who will benefit from immunotherapy,” senior author Aleksander Popel, a professor of biomedical engineering and oncology at the Johns Hopkins University School of Medicine, said in the news release.

To address this, Popel and his team applied a mathematical model known as quantitative systems pharmacology to simulate 1,635 virtual patients with metastatic triple-negative breast cancer. These simulations, involving the immunotherapy drug pembrolizumab, allowed them to test a range of biomarkers using advanced computational tools, including machine learning algorithms.

While pretreatment biomarkers derived from initial tumor biopsies or blood samples showed limited predictive capacity, on-treatment biomarkers — measurements taken after therapy began — proved more effective. Surprisingly, commonly assessed biomarkers, such as PD-L1 expression and lymphocyte presence, were better predictors of response when evaluated before treatment initiation.

Further, the team assessed less-invasive blood-based biomarkers.

“The simulated response rates increased more than twofold — from 11% to 25% — which is quite remarkable,” added Arulraj, emphasizing the potential of noninvasive biomarkers when traditional biopsy methods are impractical.

“Cancer patients can benefit tremendously from tailored treatments,” added co-author Cesar Santa-Maria, an associate professor of oncology at the Johns Hopkins Kimmel Cancer Center. “Predictive biomarkers are critical as we develop optimized strategies for triple-negative breast cancer, so as to avoid overtreatment in patients expected to do well without immunotherapy and undertreatment in those who do not respond well to immunotherapy.”

This research not only advances the field of computational oncology but also highlights the potential for similar methodologies to be applied to other cancer types — a step towards more comprehensive and personalized cancer care. Previous work by the Johns Hopkins team, which also included a computational model for late-stage breast cancer, underscores their trailblazing role in this domain.

This study’s encouraging results pave the way for future clinical trials and the potential replication of these methods for various cancers.

Ultimately, the advancements in predictive biomarker identification could markedly improve treatment outcomes, transforming the prognosis for many cancer patients and offering newfound hope for those fighting metastatic diseases.