The set cover algorithms for motif selection.

What is motif discovery?

A common biological question when dealing with ChIP-seq data is to identify transcription factor binding sites. This problem is called motif discovery. Motif discovery algorithms produce a large number of motifs (e.g., MEME) and they don’t evaluate motifs systematically. Therefore, several methods (e.g., PMID: 24335146) have been developed to select high-confidence motifs.

What is motif selection?

The motif selection problem seeks to identify a minimal set of regulatory motifs that characterize sequences of interest (e.g. ChIP-seq binding regions). The output motifs represent putative binding sites for primary transcription factors (ChIP-ed factors) and co-factors.

Getting started

Please see the documentation for more details.

The following example has been tested in Ubuntu x64.

Step 0. Download the GLPK library

https://www.gnu.org/software/glpk/

Step 1. Compile the C code

.. code:: bash

cd RILP

g++ -std=c++11 RIPL_simplex2_07_31.cc -o msdc -L/PATH/glpk/lib -I/PATH/glpk/include -lglpk

Step 2. Run the RILP method

.. code:: bash

./msdc foreground.list background.list motif_mapping.fimo

Output

The output is a list of selected motifs (i.e., subset of the input motifs) defined in motif_mapping.fimo.

Data folder

This folder contains all raw inputs (i.e., binary matrices) and the Fisher exact test p-values for every motif.

To replicate our results

Please use the detailed installation steps below

Install Anaconda Python 2.7 version

https://www.anaconda.com/distribution/

Create Conda environment

conda create -n set_cover python=2.7

conda activate set_cover or source activate set_cover

Install dependencies

• Pandas: conda install -c anaconda pandas
• Joblib: conda install -c anaconda joblib
• scikit-learn: conda install -c anaconda scikit-learn

Run the nested cross-validation framework

cd Evams

Note that Evams imports conf.py from the same directory. You can add Evams to your env path by doing export PATH=$PATH:/path_to_Evams_folder/

Evams2 -h

Evams2 -jid test -confFile example.conf

In example.conf, please change input to individual files in the Data folder.

Configuration files (.conf) used in this paper are provided in the conf_paper folder. In the same folder, running log files including CPU time and memory used are also provided.

-jid is job ID. An output folder will be created using job ID, in which all results will be put.