Overview
HubPredictor is the R function to run Bayesian Additive Regression Trees (BART) model for predicting chromatin interactions hubs using histone marks information. HubPredictor can also be used to predict topologically associated domain (TAD) boundaries.
Systems Requirements
HubPredictor was written in R language. It required the R package bartMachine, which is available in http://cran.r-project.org/web/packages/bartMachine/index.html.
Usage
Unzip the package. Change the current directory in R to the folder containing the scripts. HubPredictor.R is the R function to run BART model for predicting chromatin interactions hubs using histone marks information. HubPredictor contains two main steps: (1) read TrainingSet data & build BART regression model; (2) read input data & prediction using trained BART model. It takes three parameters: (1) file_in is the input file containing histone marks density for each chromatin anchors; (2) file_out is the output file, which generated the prediction information; (3) file_trainingset [optional] is an the dataset for training the BART model, which takes the file 'TrainingSet/Hub.txt' by default.
Example
source("HubPredictor.R")
HubPredictor(file_in='input.txt',file_out='output.txt',file_trainingset = 'TrainingSet/Hub.txt')
File Format
(1) file_in
A tab-delimited text file provides histone marks density for each chromatin anchors information, which were used as input for model prediction. Specifically, given a chromatin anchor, we summarized the local pattern for each histone mark by averaging the sequence reads over a 300kb window (about twice the average distance between an anchor and its target site) centered at the anchor location. It can be calculated using Bedtools coverage.
(2) file_out
A tab-delimited text file provides the prediciton information (in red), where group represents class predictions, and probability represents probability predictions.
(3) file_trainingset
A tab-delimited text file provides the training dataset to build BART model, which includes histone marks density of each chromatin anchor (same as file_in) and a column containing the group information. By default, the code will use the hubs defined based on the Hi-C dataset used in our paper ('TrainingSet/ Hub.txt'), which is used in our paper, to build BART model. User can define the hubs based on Hi-C dataset according to the method described in our paper. For TAD boundaries prediction, we also provided the TAD boundaries defined in the Hi-C dataset used in our paper ('TrainingSet/TADBoundary.txt'), to build BART model.
Reference
J. Huang, E. Marco, L. Pinello, G.C. Yuan. Predicting chromatin organization using histone marks. Genome Biology. 2015.
Contact
Jialiang Huang (jhuang at jimmy.harvard.edu or huangjialiangcn at gmail.com)