Speaker
Prof.
Eytan Domany
(Weizmann Institute of Science)
Description
I will present a “systems approach” to analysis of high throughput large cancer datasets. The basic idea is to make use of existing knowledge, taking the golden path between “ignorance-based” machine learning approaches and the “all details are essential” view of many biologists. This philosophy [1] has been implemented in Pathifier – an algorithm that infers pathway deregulation scores for each individual tumor sample, on the basis of expression data [2]. This score is determined in a context-specific manner for every particular data set and type of cancer that is being investigated. The algorithm transforms gene level information into pathway level information, generating a compact and biologically relevant representation of each sample. We demonstrated [2] the algorithm’s performance on three colorectal cancer datasets, two glioblastoma multiforme datasets, and on a very extensive dataset on breast cancer [3]. We showed that our multi-pathway-based representation is robust, preserves much of the original information, and allows inference of complex biologically significant knowledge. In particular, we demonstrate that one can glean clinically useful information, such as prediction of response to particular chemotherapy for a carefully selected subclass of patients. These results indicate that the prevalent search for “silver bullet” prognostic and predictive gene lists, that are supposed to work for all breast cancer subtypes, are to be replaced by much more specific (and restricted) biomarkers.
[1] Using High-Throughput Transcriptomic Data for Prognosis: A Critical Overview and Perspectives. Eytan Domany, Cancer Research 74, 4612 (2014).
[2] Pathway-based personalized analysis of cancer. Yotam Drier, Michal Sheffer, and Eytan Domany, PNAS 110, 6388 (2013)
[3] Pathway-based personalized analysis of breast cancer data. Anna Livshits et al, Molecular Oncology 9, 1471 (2015).
Primary author
Prof.
Eytan Domany
(Weizmann Institute of Science)
