TFBS Predictions

We provide two approaches to use BPAC. One is to make the prediction online, and the other is to download the program and make the prediction offline. Due to the computer power limit, the online version allows only less than 1000 candidate sites as input. Therefore, we recommend the offline version. Detailed usage instruction please see FAQ.

Online version:

We need at least two files to make prediction.
First, we need a bam file from either ATAC-Seq or Dnase-Seq. The bam file will be used to calculate the features for the prediction.
Second, we need a region of interest file with bed format ( e.g. example ), which lists the potential candidate sites.
The users can use motif scan programs, such as FIMO, to generate the candidate sites.

Reference genome: �
hg19
mm9 Please choose a reference genome

Bam file (URL of sorted alignment file): �

Please enter URL of alignment file

Is it paired end? �

Region of interest: � Please upload region file first

Name of transcription factor (TF): �

Online prediction result

Offline version:

Please download the BPAC package here. The software requires Unix/Linux environment and Python 2.
The basic workflow is:

Training

Generate features and labels → Construct model

Test

Generate features on test data → Predict using trained model → Evaluation of prediction

Installation and usage instruction is at here

BPAC: A universal model for prediction of transcription factor binding sites based on chromatin accessibility.