(2) The 1598th Biological Symposium, Friday, August 31, 2012, 10:30-11:30, Place: Seminar room - Library 3F:B301
Speaker: Vladimir Bajic (CBRC (Computational Bioscience Research Center),
KAUST (King Abudallah University of Science and Technology), Thuwal, Saudi Arabia)
Title: Transcription Factor Binding Site Models: Which One is the Best?
Summary:
Models of transcription factor binding sites (TFBSs) on DNA are frequently used in analysis of transcription regulation of both individual genes and whole genome studies. Accurate recognition of TFBSs is a challenging problem since transcription regulation is influenced by numerous factors including epigenetic and genomic characteristics, chemical properties and structures,and nucleotide interaction on long distances. While modeling of TFBSs has long history and a number of models have been reported in the literature, there was no systematic comparison of performances that different models manifest. We made a comparison study of 12 different TFBS models and estimated estimate their performance on experimental TFBSs data. The models we studied include Markov chains, hidden Markov model, Bayesian networks, thus encompassing the popular position weight matrix (PWM) model. We illustrate the efficient use of the results in discovery of prognostic biomarkers.
