Researchers at Johns Hopkins Kimmel Cancer Center and Johns Hopkins University School of Medicine have developed a novel computational method to determine which patients with metastatic triple-negative breast cancer are likely to benefit from immunotherapy. This breakthrough approach was detailed in the journal *Proceedings of the National Academy of Sciences* on October 28, underscoring the potential of predictive modeling in patient-specific treatment.

Challenges of Predictive Biomarkers in Immunotherapy

Immunotherapy, which harnesses the body’s immune system to combat cancer cells, only benefits certain patients, making it crucial to identify who will respond before treatment begins, explained lead author Dr. Theinmozhi Arulraj, a postdoctoral fellow at Johns Hopkins. "Immunotherapy can have high toxicity, so it’s vital to pinpoint patients who will benefit," Arulraj noted. Traditionally, clinicians rely on specific molecules in tumors, known as predictive biomarkers, to gauge potential treatment response. However, according to senior author Dr. Aleksander Popel, a professor of biomedical engineering and oncology, these biomarkers often lack the accuracy needed to guide treatment decisions for many patients, particularly for those with complex cancer types.

Leveraging Computational Models for Predictive Accuracy

To address this, the Johns Hopkins team utilized a quantitative systems pharmacology model to simulate 1,635 ‘virtual patients’ with metastatic triple-negative breast cancer. These virtual profiles allowed the researchers to test immunotherapy drug responses and analyze 90 different biomarkers using advanced machine learning and statistical techniques. Notably, their model highlighted two categories of biomarkers with promising predictive power: on-treatment biomarkers, measured after therapy begins, and early tumor response metrics such as changes in tumor diameter.

Interestingly, the team found that measuring changes in tumor diameter within the first two weeks of treatment could significantly enhance predictive accuracy. "This metric, which can be captured through standard CT scans, holds great potential in early identification of responders," Popel said. In a virtual trial based on early tumor diameter changes, the simulated response rates doubled, from 11% to 25%, suggesting a noninvasive way to monitor treatment efficacy without requiring invasive biopsies.

Redefining Biomarker Assessment for Broader Application

The team’s findings could redefine how biomarkers are assessed in other cancers. For instance, while certain markers like PD-L1 levels and lymphocyte presence in tumors provided valuable pretreatment data, the study found these markers were more accurate when measured before treatment rather than during. This discovery suggests that a combination of biomarker measurements taken at multiple points may be more effective in predicting outcomes.

"Using computational modeling to examine these biomarker interactions helps us understand patient-specific responses better, especially in cases where traditional methods fall short," said Dr. Cesar Santa-Maria, a breast cancer immunotherapy expert at Johns Hopkins and study co-author. Santa-Maria added that noninvasive biomarkers could serve as alternatives where biopsy collection is challenging, helping avoid overtreatment in patients likely to do well without immunotherapy.

The researchers plan to apply their findings to future clinical trials, with hopes that this method could become a model for biomarker assessment across various cancer types. Their model, built upon years of data from clinical and experimental research, was validated and developed with support from the National Institutes of Health and the National Science Foundation.