Explainable Generative Deep Learning for plant transcription factor binding sites discovery

Running the standalone program

Create & activate virtual environment and install other required dependencies

conda create -n ptfvac python=3.8.10 -y

conda activate ptfvac

unzip software.zip

chmod a+x INSTALL

./INSTALL (This step will take some time to install all dependencies in one go)

cd deeptfactor

sh meme.sh (This step will take some time to install MEME suite)

Note: Always use Alphafold2 generated TF pdb file only (An example is provided). This standalone version was developed in Linux, and has been tested in Ubuntu 20.04 with Python 3.8.10.

Input

fastafile = File containing fasta sequences.

example.pdb = Alphafold2 generated example pdb file.

Running script

To detect the TF binding sites, In parent directory execute following command:

Usage: ./ptfvac "fastafile" "folderpath" "Alphafold generated PDB file" "folderpath of deeptfactor inside PTF-Vac"

eg: ./ptfvac fastafile folderpath example.pdb /home/user/ptfvac

To generate line plot, execute the following command:

python3 plot.py plot/seq1.txt (filename) (generated plots are interactive)

For comparison between motif identified by PTF-Vac and already known motif (In MEME format, an example is provided "known.meme")

Step 1: Convert PTF-Vac motif to MEME format

python3 pwm_meme.py folderpath

Run TOMTOM

Step 2: tomtom motiflogopwm.meme known.meme -o result/folder

Output

results.txt = The TF binding site information regarding the binding site sequence and it's start and end coordinates in the sequence. The results also provides the binding site sequence importance score, its binomial test based p-value, and hypergeometric test based p-value.

motiflogopwm.csv = Position weight matrix of binding site sequences.

motiflogopwm.pdf = Sequence logo of binding site sequences.

plots = A folder containing importance score plot.

Note

For HADDOCK implementation see README_HADDOCK.md file for more details.