Novel approach uncovers possible drugs based on specific cellular targets for treating glioblastoma

Researchers have developed a new computational approach that uncovers possible drugs for specific cellular targets for treating glioblastoma, a lethal brain tumor. This approach enabled them to predict more effective treatment combinations to fight the disease on an individualized basis. This laboratory and computational research effort was led by scientists at Georgetown's Lombardi Comprehensive Cancer Center. 

The cellular targets we identified could be key to effectively fighting a disease that has seen only one new targeted drug approved in the last two decades."

Nagi G. Ayad, PhD, senior author, associate director for translational research, and professor of oncology at Georgetown Lombardi

"Although this research focused on glioblastoma, we hypothesize that our framework and algorithms will be useful for many other cancers and diseases," adds Robert K. Suter, PhD, lead author and assistant professor, also at Georgetown Lombardi.

The finding appears Month Date, 2026, in Nature Communications.

In glioblastoma and other cancers, types of tumor cells vary dramatically even within a small tumor. The tumor cells are constantly adapting by turning genes on, off, or somewhere in between to suit the environment around them. These variable cell states represent moving targets in the context of treating these cancers, which have made them highly intractable.

Glioblastoma is the most common malignant adult brain cancer, with a median overall survival of only 15 months. Over 10,000 people a year die from this disease in the U.S., with only 7 percent of people with glioblastoma living more than 5 years after diagnosis.

In order for the researchers to be able to identify effective drugs against the heterogenous and constantly morphing cancer cell mixture in glioblastoma, they needed to develop a platform that could predict the differing sensitivity and resistance of diverse cell types to various possible treatments.

To this end, they created scFOCAL (Single-Cell Framework for -Omics Connectivity and Analysis via L1000), a program that uses RNA sequencing information from individual cells of newly diagnosed and recurrent glioblastoma tumors to predict how these tumors will respond to different treatments. They were able to identify compounds that oppose the gene expression signatures of distinct glioblastoma cell states and then leverage this capability to predict combinations of drugs that might work well together to target the variable tumor cell landscape.

Looking at RNA instead of DNA was one key to their discovery. Knowing the transcriptional profile, or RNA levels in a tumor cell, allowed the researchers to better understand what a cell was doing at any given moment, and to make predictions about which drugs can be used at a specific time. It is much harder to do that using information based on DNA, says Ayad, as DNA changes very little over time.

For their analyses, the researchers were able to pull information from a large drug repository, called the NIH LINCS (Library for Integrated Network-based Cellular Signatures) L1000, and use this information to derive gene expression signatures for each small molecule in the dataset. While the LINCS L1000 small molecule library is not all-encompassing, the researchers note that they can extrapolate from molecules that are prioritized in a library search to identify molecules of interest with similar properties and then synthesize new drugs based on those properties.

"We are expanding our framework and research to be able to make predictions about how to manipulate multiple cell types. We want to identify small molecules that push other cell types that are active within tumors to more favorable gene expression states," says Suter. He also notes that in the future, a framework like scFOCAL could be used to identify sequences of treatments, where one drug is used first, the tumor responds and assumes a new state, which would indicate a second or even third sequence of treatments to keep up with the changing landscape of the tumor.

"This work has benefited greatly from close collaborations with colleagues who are neuro-oncologists and neurosurgeons. Several of these colleagues have suggested designing clinical trials based on our findings. The more data we have from patients, the more robust our predictions will be, so these collaborations will be key to the design of future clinical trials," concludes Ayad.

Ayad and Suter are the co-creators of copyright-protected software and algorithms owned by Georgetown University related to the technology that is described. The other co-authors report no other related financial interests.

In addition to Ayad and Suter, the other authors at Lombardi include Anna M. Jermakowicz, Rithvik Veeramachaneni, Matthew D'Antuono, Longwei Zhang, Rishika Chowdary, Simon Kaeppeli, Madison Sharp, Grace Baker, Luz Ruiz and Pravallika Palwai; Vasileios Stathias, Winston Walters, Maria Cepero, Sion L. Williams, Michael E. Ivan, Ricardo J. Komotar, Macarena I. De La Fuente and Stephan C. Schürer are at the University of Miami Miller School of Medicine, Miami; Danielle Burgenske and Jann N. Sarkaria are at the Mayo Clinic, Rochester, MN. Edward B Reilly, Anatol Oleksijew and Mark G. Anderson are at AbbVie, Oncology Discovery, North Chicago, IL; Gregory Stein is at Curtana Pharmaceuticals, Inc., Austin, TX; and Alexandre Wojcinski and Santosh Kesari are at Pacific Neuroscience Institute and Saint John's Cancer Institute, Providence Health System, Santa Monica, CA.

This work was supported by NIH grants (RM1NS13303, P30CA051008, P30CA240139 and U54HL127624). Additional support came from Bellringer as well as an American Cancer Society Institutional Research grant IRG-23-1156148-27-IRG. This work used Jetstream2 at Indiana University, which is supported by National Science Foundation grants (2138259, 2138286, 2138307, 2137603 and 2138296).