SieveAI: An Automated Drug Discovery Pipeline

SieveAI: An Automated Drug Discovery Pipeline

Vishal Kumar Sahu1 , Isha Zafar1 , Shuchi Nagar2 , Amit Ranjan1# , Soumya Basu1#

  1. Cancer and Translational Research Centre, Dr. D. Y. Patil Biotechnology and Bioinformatics Institute, Tathawade, Pune – 411033
  2. Bioinformatics Centre, Dr. D. Y. Patil Biotechnology and Bioinformatics Institute, Tathawade, Pune – 411033

Corresponding Authors

  • Dr. Soumya Basu (soumya.basu@dpu.edu.in)
  • Dr. Amit Ranjan (amith.ranjan@dpu.edu.in)

Background

Computational approach of molecular docking and dynamics has played a crucial role in finding and understanding therapeutic agents and targets in various diseases and altered metabolic bioprocesses. Over time the increase in the information of molecular structures has led to enhanced computational algorithms and more complex workflow. So the manual approach of drug discovery becomes time consuming and error prone. Though there are packages which have partial automated workflow in the public domain but no package available to automate the complete docking analysis. In the presented pipeline, SieveAI (version 0.4), we have created an automated docking and analysis pipeline using python having compatibility for Windows and Linux. It can be extended and exploited to use multiple docking packages and web based tools. The pipeline currently involves usage of open source tools for automation. In its current version, it can perform blind docking and site specific docking by processing and preparing molecular structures for docking using VINA and analysis using ChimeraX. It is suitable for studying the interaction of a single drug on multiple proteins/nucleic acids, multiple drugs with single protein/nucleic acids, or multiple drugs with multiple proteins/nucleic acids. It can be extended to automate web server based docking, prediction, and analysis tools. As a benchmark for the performance, it can dock 1000 complexes in 24 hours on Intel i5 with hexa-core CPU and 8 GB RAM on Ubuntu. The pipeline is under the process of copyright.

Automated drug discovery pipeline: 

  1. Processing and analysis of molecular interactions involving protein-protein, protein-small molecule, protein-RNA, protein-DNA, RNA-RNA, DNA-RNA
  2. Interaction analysis like hydrogen bonds and contacts
  3. Ranking based on the docking score
  4. Can be extended to rescore or dock with multiple docking programs
  5. Can be used as one command solution

Conclusion

The pipeline can be helpful in automating the manual workflow for molecular docking and drug discovery including rescoring and analysis of molecular interactions. Manual workload and human errors can be overcome.

Availability

  1. PyPi.org
  2. GitHub

Future Prospects

  • Development of Graphical User Interface
  • Hosting on web-server for public use