Glioblastoma is the most common and deadly primary brain cancer in adults, and treatment options have changed very little over the past two decades. One of the biggest challenges in treating this disease is intratumor heterogeneity. Even within a single tumor, cancer cells exist in many different transcriptional states, meaning they behave differently and respond unevenly to therapy. This diversity makes it difficult to find treatments that are effective across the entire tumor.

Researchers at Georgetown University’s Lombardi Comprehensive Cancer Center developed a new computational framework to address this problem. The method, called ISOSCELES, integrates single cell RNA sequencing data from glioblastoma tumors with large drug response datasets to predict which cancer cell states are sensitive or resistant to specific therapies.

Single cell RNA sequencing plays a key role in this approach by capturing what individual tumor cells are doing at a given moment. Unlike DNA, which changes slowly, RNA reflects the current functional state of a cell. By analyzing RNA profiles from newly diagnosed and recurrent glioblastoma tumors, the researchers were able to define distinct transcriptional states within tumors and match those states to drug response signatures from the LINCS L1000 database.

Using ISOSCELES, the team identified compounds whose transcriptional effects oppose the gene expression programs of specific glioblastoma cell states. This allowed them to predict which drugs might be effective against certain populations of tumor cells, and which cells might resist treatment. Importantly, these predictions were validated across multiple experimental systems, including cell culture, tumor slices, and animal models.

One notable outcome of the work was the identification of a promising drug combination targeting glioblastoma. By pairing an OLIG2 inhibitor with Depatux M, the researchers found enhanced activity against tumor cell populations that would likely survive single agent treatment. This supports the idea that combination therapies guided by transcriptional state may be necessary to overcome tumor heterogeneity.

Overall, ISOSCELES provides a framework for translating single cell RNA sequencing data into actionable therapeutic insights. By predicting drug sensitivity and resistance across diverse tumor cell states, this approach could help guide the development of more effective combination treatments for glioblastoma and potentially other highly heterogeneous cancers.

Availability – The ISOSCELES package source code is available at https://github.com/AyadLab/ISOSCELES,