Glioblastoma multiforme (GBM) is a highly aggressive form of brain cancer, with a median survival time of just one year following diagnosis. Treatment is complicated by the considerable variability in GBM tumors – what works for one tumor often fails to work for another. Selecting the most appropriate treatment may soon, however, be easier, thanks to an improved GBM classification system, developed by Mount Sinai and Sema4 scientists and published in Cancer Research.
A team of researchers, led by Sema4’s Head of Data Sciences Jun Zhu, PhD, reasoned that an improved GBM classification system could help clinicians to select the most pertinent therapy – a case of “know your enemy”. Some GBM tumors are dependent on the mitotic spindle checkpoint molecule BUB1B for their survival, so his team mined complex datasets to produce an innovative computational method to classify tumors based on their BUB1B dependency. In doing so, they uncovered new tumor subtypes and found that while BUB1B-sensitive tumors had a significantly worse prognosis, they were also predicted to be more responsive to many of the cancer drugs already in clinical use.
The molecular subtypes identified in this new study appear to provide a more accurate estimate of prognosis and therapeutic response than existing classifications. One reason that previous classifications have failed to lead to effective personalized treatments is the high degree of intratumoral heterogeneity in GBM. Cells from different parts of the tumor may belong to different molecular subtypes and, therefore, subtype-specific therapies fail to eradicate all the cancerous cells. The BUB1B classification system, however, does not suffer the same defect.
“It was a pleasant surprise to us that our subtype is stable for heterogeneous tumor cells within a GBM tumor and, thus, it is possible to kill all tumor cells instead of just a subgroup,” says Dr. Zhu. “Preliminary results indicate that the stability is associated with certain genomic features, but more data are needed to understand why. More importantly, we also need to work out how to leverage the subtype information to develop mechanism-specific therapies.”
“These findings underscore the significant potential we see to improve patient outcomes by investing in predictive modeling of even the most complex types of cancer,” explains Eric Schadt, PhD, Sema4’s CEO and Dean for Precision Medicine at Icahn School of Medicine at Mount Sinai.
The study was the result of a multidisciplinary collaboration between computational scientists and clinicians – a characteristic of many Sema4 research projects. Information generated from our integrative studies – such as the GBM project and a recent examination of lung cancer mutations – is the first step towards designing improved diagnostic tests and optimizing personalized cancer therapies. Currently, Sema4 offers the Oncology Hotspot Panel, which provides information on over 200 mutational hotspots associated with a range of cancers. As our knowledge of cancer genomics increases so too will our ability to expand this number, leading to improved diagnosis, treatment, and survival rates for cancers including glioblastoma. “We look forward to building on this collaborative project and moving toward development of a diagnostic test that could help physicians better understand and treat their patients’ glioblastoma cases,” says Dr. Schadt.