SUpervised PRomoter REcognition Framework

For research on promoter recognition using deep learning.

This project is being reworked to reduce resource utilisation (Mostly RAM usage). Updates are introduced into suprref folder. Previous version including previously published models and data are available in the previous folder. Documentation for the previous version at https://ivanpmartell.github.io/masters-thesis/, Analysis tools also available in the analysis folder.

Docker installation

Requires docker or docker desktop.

Windows:

build-docker.bat

Then open a command prompt (cmd.exe) and run the executable:

suprref.bat

Linux/Unix:

build-docker.sh

Then open a terminal and run the executable:

suprref.sh

Manual Installation

Requires python >3.6.9 and pip >21. First create the local wheel:

python setup.py bdist_wheel

Then install it:

pip install dist/suprref-0.1-py3-none-any.whl

This will create

This should automatically install any required python libraries that SUPRREF uses, including:

• numpy
• pytorch
• tqdm
• skorch
• pandas
• biopython
• pyfaidx
• PyInquirer
• pyfiglet
• mysql-connector-python

Usage

After installation you can run the executable with arguments.

• Bring up the help manual suprref -h

Data

Data consolidated in previous/data/ folder is obtained through UCSC, EPD, MGA, as well as resources from other promoter recognition researchers. Certain folders inside the data folder contain a readme file to attribute ownership of data within them. Other data files that are too big for the repository can be obtained through bash scripts, e.g. human chromosomes. Data shown in our results and tables can be obtained using the bash scripts inside the previous/ folder.

Analysis and visualisation

Visualisation and analysis of promoter files was done using jupyter notebooks found in analysis/models/notebooks/. Further analysis and visuals can be found in the PowerBI file at previous/ folder.

Source code

Code for the command-line tool can be found inside the suprref/ folder, while code for the cross-testing, training and testing of the reimplemented literature model can be found inside the previous/ folder.

Troubleshooting

This python framework has been tested on Ubuntu 18.04 (bionic) using the python3 and pip3 commands found in the default repositories.

Remember to run all commands inside the promoters (this repository) folder.

If you do not obtain the suprref executable with the manual installation, try using a virtual environment:

• Install virtualenv for the creation of virtual environments pip install virtualenv

• Create the virtual environment virtualenv venv

• Activate the virtual environment source venv/bin/activate

And follow the manual installation instructions.