tfbsincbeta: Software for a Bayesian method for estimating the number of binding sites for a transcription factor (TF), based on a known position-frequency matrix (PFM) for the TF, within a set of promoter sequences. The method incorporates a prior probability distribution on the number of transcription factor binding sites (TFBS) that is similar to the beta-binomial distribution but with modifications to account for the double-stranded nature of DNA. Samples from the posterior probability distribution of the number of TFBS are generated using a Metropolis-Hastings algorithm with a proposal generator that is weighted based on the Shannon entropies of the probabilities for presence/absence of a binding site each possible TFBS position. The software accompanies the manuscript "An empirical prior improves accuracy for Bayesian estimation of transcription factor binding site frequencies within gene promoters" by Stephen Ramsey, which has been submitted to the journal Bioinformatics and Biology Insights.
Author: Stephen Ramsey, Oregon State University (lab.saramsey.org)
Date: Sept. 10, 2016
License: This software is distributed under the Apache Software License 2.0. Please see the file LICENSE for details on the software licensing agreement.
Usage notes: The R script, "tfbsincbeta.R", reads a data file "Matrices.txt" of TFBS PFMs in tab-delimited format (see header comment for "tfbsincbeta.R" and an example file in "data"). In order to generate the empirical performance results in the above-referenced article, the "Matrices.txt" file contains all TF PFMs from the TRANSFAC Professional database version 2015.1. That database PFMs can be obtained from QIAGEN but the author is not permitted to redistribute the database. The R script makes use of the R package "parallel".