From Promoter to Expression – A Probabilistic Framework for Inferring Regulatory Mechanisms

January 31, 2005
Yoseph Barash | Hebrew University

Inferring regulatory mechanisms based on in silico analysis of regulatory elements has been the target of much research efforts in recent years. Specific aims include identifying combinatorial interactions of either known or novel transcription factors; identifying the cellular conditions in which these combinations are activated and which genes they control; pinpointing the active binding sites of each transcription factor; and finally predicting the expression profile of a gene given its promoter sequence. I will present a sequence of works that use a flexible probabilistic framework to handle these tasks. The framework we developed is based on probabilistic graphical models, such as Bayesian Networks. Using these models we aim to combine diverse sources of genomic data in a synergistic manner and then use the learned models to generate biological hypothesis. I will demonstrate the possible gains from these models when analyzing transcriptions factors from the TRANSFAC data base and when analyzing the Yeast genome combined with various high throughput data.

Speaker Details

Yoseph Barash received a B.Sc. in Computer Science & Physics from the Hebrew University in 1999. Since then he has been a Ph.D. student in Prof. Nir Friedman’s group at the Hebrew University. His main research interest is in developing machine learning based algorithms to solve problems from the biological/ medical domains. He focuses on building probabilistic graphical models for regulatory mechanisms using diverse sources of data such as genomic sequence, gene expression and location (ChIP) measurements.