Bayesian Identification of Transcriptional Regulators from Epigenomics-Based Query Region Sets

BIT is a novel bioinformatics tool designed to identify transcriptional regulators from a user-provided set of epigenomic regions. These regions are typically derived from genome-wide epigenomic profiling data that probe specific biological processes, such as ATAC-seq. Using a pre-processed reference library containing over 10,000 human TR ChIP-seq datasets and over 5,000 mouse TR ChIP-seq datasets, BIT leverages a Bayesian hierarchical model to achieve higher accuracy and enhanced interpretability.

Read the Docs Documentation

We have an manual on Read the Docs.

Manual

In this manual, we provide an introduction to BIT, a detailed walkthrough of BIT's main installation and primary functions, and four examples to reproduce figures from our manuscript.

Reference ChIP-seq Data

We have pre-compiled 10,140 TR ChIP-seq datasets associated with 988 human TRs and 5,681 TR ChIP-seq datasets associated with 607 mouse TRs. We also offer different bin-width options for the TR reference database to suit user needs. These data can be downloaded from the Zenodo online data repository:

Bin Width	hg38 Download Link	mm10 Download Link
1000	hg38_1000.tar.gz	mm10_1000.tar.gz
500	hg38_500.tar.gz	mm10_500.tar.gz
200	hg38_200.tar.gz	mm10_200.tar.gz

Please note: BIT cannot run without the reference data, so please load the reference ChIP-seq database once BIT has been installed.

Installation

We provide an online portal: <Online Portal>

You can also install the development version of BIT from GitHub with:

# install.packages("devtools")
#devtools::install_github("ZeyuL01/BIT")
#library(BIT)
#> ℹ Loading BIT

If you meet a compilation problem,

For Mac users please refer to the following: https://thecoatlessprofessor.com/programming/cpp/r-compiler-tools-for-rcpp-on-macos/

For Windows users please refer to the following: https://cran.r-project.org/bin/windows/base/howto-R-devel.html

Or submit your questions through issues, we are happy to answer them.

Load ChIP-seq Data

In this example, we use a bin-width of 1000. For data with other bin-widths, please refer to the pre-compiled ChIP-seq database available on Zenodo.

# Download the ChIP-seq file from Zenodo and unzip it to a local directory.
# For example, unzip to: /.../Desktop/data/ChIP-seq/hg38/

# Set the path to your unzipped ChIP-seq data.
chip_path <- "/.../Desktop/data/ChIP-seq/hg38/"

# Set genome = "hg38" for hg38.zip and "mm10" for mm10.zip.
# Run the function below once to load the data.
load_chip_data(chip_path, bin_width = 1000, genome = "hg38")

Quick Start to Use BIT

Running BIT is very simple as long as your input file is in one of the supported formats: bed, narrowPeak, broadPeak, bigNarrowPeak, or csv. Note that if your input bed file includes a column named "summit", those values will be used instead of the region's midpoint when converting to binarized vectors.

#As an example, the user-input peak set file path is "file_path/CTCF.bed"
input_path = "file_path/CTCF.bed"
output_path = "output_path/"
BIT(input_path, output_path, N=5000, burnin=2500, genome="hg38")
#> Loading and mapping peaks to bins...
#> Done loading.
#> Comparing the input regions with the pre-compiled reference ChIP-seq data, using a bin width of 1000 bps...
#> Loading meta table...
#> Starting alignment process...
#>   |==================================================| 100%
#> Alignment complete.
#> Starting BIT Gibbs sampler with 5000 iterations...
#> 0%   10   20   30   40   50   60   70   80   90   100%
#> [----|----|----|----|----|----|----|----|----|----|
#> **************************************************|
#> Gibbs sampling completed.
#> Output data saved as output_path/CTCF.rds
#> Loading data from file...
#> Processing theta matrix and TR names...
#> Compiling results...
#> Results saved to output_path/CTCF_rank_table.csv
#> BIT process completed.


#> CTCF_results<-read.csv(paste0(output_path,"CTCF_rank_table.csv"))
#> CTCF_results
#         TR   Theta_i     lower     upper  BIT_score BIT_score_lower BIT_score_upper Rank
#1      CTCF -2.010571 -2.011593 -2.009676 0.11809745      0.11799114      0.11819079    1
#2     RAD21 -2.028610 -2.031747 -2.025619 0.11623164      0.11590978      0.11653925    2
#3      SMC3 -2.110100 -2.120542 -2.100907 0.10811898      0.10711622      0.10900866    3
#4     SMC1A -2.181305 -2.193447 -2.170246 0.10144192      0.10034048      0.10245443    4
#5      PHF2 -2.222166 -2.633869 -1.990377 0.09777755      0.06699025      0.12021697    5
#6    GABPB1 -2.301673 -2.689461 -2.062208 0.09098450      0.06359813      0.11282466    6
#7      CHD2 -2.317842 -2.370804 -2.270597 0.08965600      0.08542628      0.09358759    7
#8      NFYA -2.353494 -2.396030 -2.313238 0.08678843      0.08347594      0.09003250    8
#9    NELFCD -2.371794 -2.476460 -2.276018 0.08534898      0.07752502      0.09312872    9
#10     NFYC -2.373795 -2.767204 -2.122314 0.08519290      0.05912233      0.10694685   10

Additional examples

Runx1 Knockout dataset.

scPerturbation dataset.

Cancer dataset.

PBMCs dataset.

Hardware Overview (Mac and HPC)

Below is an overview of the system hardware information (Personal, Macbook Pro):

Attribute	Value
Model Name	MacBook Pro
Model Identifier	MacBookPro18,3
Model Number	MKGQ3LL/A
Chip	Apple M1 Pro
Total Number of Cores	10 (8 performance and 2 efficiency)
Memory	16 GB

For HPC information, please refer to the SMU M3 manual: SMU M3 Manual

Appendix

If you use BIT in your work, please cite us!

Lu, Z., Xu, L. & Wang, X. BIT: Bayesian Identification of Transcriptional regulators from epigenomics-based query region sets. Nat Commun 16, 4966 (2025). https://doi.org/10.1038/s41467-025-60269-4

Dr. Xinlei (Sherry) Wang Dr. Lin Xu