FFFold is a web application for the local optimisation of protein structures predicted by the AlphaFold2 algorithm and deposited in the AlphaFoldDB database. Protein regions predicted with confidence less than 90 are optimized by the physics-based generic force field GFN-FF accelerated by a divide-and-conquer approach which results are comparable to the optimisation of whole protein structure with constrained α-carbons. Thus, FFFold optimises in particular the bond lengths and angles and describes the interactions between nearby residues. Before computation of the charges, input protein structures are protonated by PROPKA3. The details about the methodology and usage are described in the manual. This website is free and open to all users and there is no login requirement.
To run FFFold locally, you will need to have Python 3.11 and pip installed. In addition, optimisation software xtb is needed, which can be obtained via Miniconda.
Then, clone the project and install the project dependencies by running:
$ cd /opt
$ git clone --recurse-submodules --depth 1 https://github.com/sb-ncbr/FFFold
$ sudo python3.9 -m venv venv
$ . venv/bin/activate
$ pip install -r requirements.txtRun the project by running the following command inside the virtual environment:
(venv) $ cd /opt/FFFold/app
(venv) $ export FLASK_APP=routes.py
(venv) $ flask runand point your browser to localhost:5000/.
MIT