Ofer Lavi

MSc student at the School of Computer Science at the Tel Aviv University, under the supervision of Prof. Ron Shamir.

Lab phone: +972-3-640-5394

E-mail: 999oferlav1@tau.ac.il999 (remove the 999 unless you are into spamming...)

Phone: +972-3-640-5394

My Msc. Thesis Research

Analysis of gene expression using microarrays went into clinical use in the recent years after FDA approval for prediction of effectiveness of chemotherapy against various cancers. Prediction methods are based on a small set of genes, specifically selected in a computational or statistical manner and show significantly higher success rates compared to previous, more traditional methods.

However, these methods have drawn critique from both biomedical and computational scientists. For the life-scientists, the fact that treatment decisions are made as a “black box” fashion, treating genes as numbers, and lacking any biological and cellular rationale is dissatisfying, which we call the Blind Analysis Disbelief. From a computational perspective, these methods suffer from shortage of training data: thousands of expression values were measured for each patient, but only in a few hundred patients, a problem coined by the late Richard Bellman as the Curse of Dimenstionality

We are using additional information to answer the shortcomings of the current methds for selecting diagnostic markers. By collecting published and experimental information, many biological systems can be summarized as signaling networks. We combine expression profiles of cases and controls with large scale protein-interaction networks in order to find better diagnostic markers. These markers are better in terms of their prediction power, and are more readily interpretable in terms of biological processes involved. As such, they better understood by medical practitioners and are also useful for downstream research.