Pathway-based personalized analysis of cancer Author Interview:
Prof. Eytan Domany

Department of Physics of Complex Systems and Department of Biological Regulation, Weizmann Institute of Science, Rehovot, 76100, IsraelDepartment of Physics of Complex Systems and Department of Biological Regulation, Weizmann Institute of Science, Rehovot, 76100, Israel What are the main findings of the study?

Prof. Domany: The findings are two-fold: methodological and clinical.  A novel method was introduced for personalized analysis of cancer, and was applied on large colon cancer and glioblastoma datasets.

The method uses high throughput (gene expression) data to infer a pathway deregulation score (PDS) for individual tumors, for hundreds of pathways and biological processes. The method is knowledge-based in that it uses well known information about the assignment of genes to biologically relevant pathways. No detailed knowledge of the underlying networks of interactions and activations is necessary. Each tumor is represented by a few hundred of these PDSs, and further analysis uses this representation.

One of the clinically relevant findings is the discovery that the reported relatively longer survival of subjects with neural and proneural glioblastoma is due to a new subtype of these tumors – when these are excluded, the neural/proneural patients do not survive longer than the other subgroups. For both diseases, pathways whose deregulation level is indicative of prognosis were discovered and validated on independent datasets.  What should clinicians and patients take away from your report?

Prof. Domany: This is a new and promising way to use high-throughput genomic data for prognosis, and hopefully also for personalized prediction of response to therapy. Even though there is still a long way to go till direct clinical applicability of the method, some of the findings are very promising. The reported sub-stratification of glioblastoma patients provides a robust prognostic predictor. In colon cancer two pathways were found with deregulation scores that exhibit very significant correlation with survival: CXCR3-mediated signaling and oxidative phosphorylation – both may find their way into the clinic as prognostic tools. What recommendations do you have for future research as a result of this study?

Prof. Domany: On the methodological side, this study demonstrates how tumors can be represented by individual “higher level” biologically relevant scores. On a more fundamental level, the study proves the strengths and advantages of a phenomenological approach, taking a golden path between ignorance-based machine learning approaches and the overkill of requiring full knowledge of every mechanistic detail. As to resulting clinical research, once the clinical relevance of the deregulation score of a pathway is substantiated, its level of deregulation in a particular tumor sample may be assessed directly, with no need for measurement of expression. Such direct measurements of pathway activity can then be used as a reliable and robust personalized prognostic biomarker.


Pathway-based personalized analysis of cancer

Yotam Drier, Michal Sheffer, and Eytan Domany
PNAS 2013 ; published ahead of print April 1, 2013, doi:10.1073/pnas.1219651110


Last Updated on March 19, 2014 by Marie Benz MD FAAD