A hidden Markov model-based method for transcription factor binding site prediction
Speaker: Voichita D. Marinescu , Children
Date: November 21 2005
Time: 11:30AM to 1:00PM
Location: TOC Lab 32-G575
Host: P Clote/ BC & B Berger/ MIT
Contact: Kathleen Dickey, 617 253 3037, kvdickey@mit.edu
Relevant URL: http://www-math.mit.edu/compbiosem/****in 10 minutes****
Computational prediction of transcription factor binding sites (TFBSs) in DNA sequences is challenging for several reasons. First, the models that abstract TFBS characteristics have to be trained on short nucleotide sequences usually between 6 to 20 bp in length that were determined experimentally and, therefore, are available in a small number. Secondly, searching long DNA sequences for matches to these models produces a large number of false positives that are difficult to separate from the true positives based on computational criteria or to evaluate experimentally in a large-scale manner. In this talk we present a method for TFBS prediction based on hidden Markov models that used the HMMER software package to build a large library of models for hundreds of transcription factors starting from sequences of experimentally determined sites curated in the TRANSFAC and JASPAR databases. This method served as the basis for developing MAPPER, a modular platform for TFBS identification and analysis, publicly available at http://mapper.chip.org/, that currently includes a database of putative TFBSs found in all upstream sequences of the human, mouse and Drosophila genomes, as well as a search engine to scan any DNA sequence or gene from the yeast, worm, fly, mouse or human genomes. We will present the results of an extensive evaluation performed on a collection of experimentally determined binding sites as well as on synthetic data that allowed us to asses the ability of our method to identify true positives, to estimate the proportion of false positives returned and to compare its sensitivity and specificity with other similar tools available. We will present several biological applications of this method and we will discuss further improvements of the HMMER modeling procedure designed to increase the sensitivity and specificity of the resulting TFBS models.
MIT
Department of Mathematics
& The Theory of
Computation Group
At CSAIL
See other events that are part of Bioinformatics Seminar Series 2005/2006
See other events happening in November 2005
