Leveraging “Digital Twins” to Predict Treatment Response in Brain Cancer

Despite decades of advancements in surgical techniques, chemotherapy, and radiotherapy, the clinical outlook for patients diagnosed with primary brain tumors remains sobering. While glioblastoma and other high-grade gliomas are relatively rare compared with other malignancies, they represent a disproportionate burden of cancer-related mortality. According to researchers at the University of Michigan, the lack of significant progress in survival outcomes necessitates a paradigm shift toward highly personalized, data-driven therapeutic strategies.

In a recent interview regarding a breakthrough study published in Nature Communications, Baharan Meghdadi, PhD, co-first author of the study, detailed the development of a "digital twin" platform designed to simulate tumor metabolism and predict how individual patients might respond to specific targeted therapies.

The Stagnation of Standard Care

The motivation behind this line of research stems from a persistent lack of improvement in the standard of care.

"Brain cancer... has been among the rare cancer types, but if you look at the death rates, it’s among the top 10 cancers in humans," Meghdadi noted. She highlighted a concerning epidemiological trend: "Over the past 2 decades, the mortality rate in the US doesn’t change much. Considering the standard of care, which is resection and chemotherapy [and] radiotherapy that has been done in the past decade, we see almost no change in mortality rate."

This plateau suggests that the "one-size-fits-all" approach of the Stupp protocol has reached its ceiling. To move the needle on survival, the research focuses on the metabolic heterogeneity of tumors.

"We need to do more research about the kind of personalized treatments that we can do for each patient to improve outcomes,” Meghdadi explained.

Mapping the Metabolomic "Cell Business"

The core of the digital twin technology lies in metabolomics—the large-scale study of small molecules within cells. By understanding the specific "business," as Meghdadi puts it, of a cancer cell, researchers can identify vulnerabilities that are absent in healthy tissue.

The process begins with the infusion of labeled nutrients.

“We infuse patients with a heavier form of glucose, which is going to be consumed in the cells, and it produced other metabolites that do some functions in the cell,” Meghdadi explained.

By tracking these heavy isotopes, the team can map metabolic flux: the rate at which molecules move through metabolic pathways. This differentiation is critical.

"By actually measuring the heavy mass of these metabolites, we can then find out what pathways... are important in the cancer, compared [with] the normal tissue," Meghdadi said. "These pathways or the conversion of these metabolites to each other are different in cancer cells compared to the normal cells. They need to alter their metabolism to proliferate."

The Digital Twin as a Computational Clinical Trial

The complexity of these metabolic networks makes it impossible to predict drug responses through simple observation. This is where the "digital twin" comes into play. By integrating isotopic carbon measurements with first-principle–based modeling and artificial intelligence (AI), the team can create a computational replica of a patient’s specific tumor metabolism.

"The digital twin approach that we used here is simulating the metabolism in the tumor cells and normal cells, because we cannot measure the important metabolic pathways in real patients," Meghdadi explained. "So, by using a digital twin, it’s kind of a computational clinical trial."

This approach allows for a "fail-fast" environment where multiple therapeutic agents can be tested before they ever reach the patient.

"Before we actually use any kind of drug in a real patient, we can predict how it can respond in the in a patient by doing this kind of principle modeling... and using AI and machine learning, so that gives us some predictions ahead of doing it in a clinical trial," she added.

Targeting the Building Blocks of Malignancy

The ultimate goal of this metabolic mapping is to identify drugs that can "starve" the cancer while sparing the brain’s healthy parenchyma. By estimating the conversion rates of metabolites, researchers can pinpoint which pathways the cancer relies on to build its structural components.

"We can then find out which pathways are more important to the cancer cells, making the building blocks of their DNA and RNA to proliferate," Meghdadi noted. "And so, if there is a drug that can target those pathways, we can use them for that patient to kill the cancer cells, and it doesn’t harm their normal cells, because they don’t rely much on that pathway."

Future Directions: Moving Toward the Clinic

While the platform has shown significant promise in preclinical settings, the transition to bedside application remains the next major hurdle. Researchers have already validated the model's predictive power in animal studies. However, the team is cautious about the leap to human application, emphasizing that prospective validation is essential.

"In the future, in order to see if this method can work, we need to do some kind of clinical trial in patients,” she concluded.

As the oncology community continues to grapple with the recalcitrance of glioblastoma, the integration of metabolomics and AI-driven digital twins offers a potential pathway toward the long-sought goal of truly individualized neuro-oncology.

REFERENCES

1. Meghdadi B, Al-Holou WN, Scott AJ, et al. Digital twins for in vivo metabolic flux estimations in patients with brain cancer. Cell Metab. 2026 Jan 6;38(1):228-246.e17. doi: 10.1016/j.cmet.2025.10.022. Epub 2025 Dec 1. PMID: 41330373; PMCID: PMC12695069.

2. Brain cancer digital twin predicts treatment outcomes. News release. University of Michigan. January 12, 2026. Accessed March 4, 2026. https://tinyurl.com/3p4mwpn6