Graph Neural Network Model to improve docking predictions

DockBox2 (DBX2) is a sequel to DockBox that combines the concept of consensus docking with machine learning to improve docking predictions. In short, DBX2 provides the ability to train and run a GNN model based on inductive representation learning (GraphSAGE) to better interpret docking results (e.g., generated by DBX). DBX2 can be used in two modes: the 'node' mode which estimates pose correctness, and the 'graph' mode, which estimates binding affinity.

The easiest way to install DockBox2 is to create a virtual environment. In this way, DockBox2 and its dependencies can easily be installed in user-space without clashing with potentially incompatible system-wise packages.

Once virtualenv has been properly installed, simply type (and press the return key)

virtualenv env

on the command line followed by

source env/bin/activate

to activate the virtual environment (do not forget to activate your environment every time you log into a new shell environment).

Finally, the DockBox2 package can be set up by going in DockBox2 installation directory and typing:

python setup.py install

Installation is complete!