Molecular docking is an essential process in scientific drug discovery. By using the recent advances in generative deep learning the runtime required for such docking studies can be reduced significantly without comprimising accuracy compared to traditional methods.
In this article, let us get familiarize ourselves with one open-source molecular docking codebase. The codebase named DiffDock is based on a recent (2022) arxiv paper. For a task of protein-ligand molecular docking, the tool can be an alternative to costly molecular docking softwares.
Setting it up in a local computer
DiffDock being an open-source tool, bioinformaticians need not worry about commercial license and the tool can be setup in any Windows or Linux machine. The setup instructions are available in the GitHub repo.
Interactive Online Tool
The cool thing about DiffDock is that it is available as an interactive online tool in Hugging face platform. This requires no setup from the user and can be readily used.
DiffDock Colab Notebook
Moreover, there is a Google Colab notebook available which takes care about everything including setting up the tool in Colab. The user can give the PDB ID of the protein and SMILES or PubChem ID of the ligand they want to do the docking studies. The docking results can be visualized in the notebook itself or can be downloaded.
By empirical studies, the authors in the paper shows that DiffDock outperforms the state-of-the-art techniques on protein binding tasks. Having fast inference times and better confidence estimates DiffDock provides a good alternative for bioinformaticians to try.
Good👍🏻