Title : Structure elucidation of drug targets through homology modeling in the AI era
Abstract:
Target/Structure-based drug design needs the presence of the target structure to move forward in the drug discovery process. However, the number of macromolecule structures available in the available data banks is limited relative to the high macromolecule sequence release. Homology modeling comes as a crucial rescue to solve this challenge. Homology modeling is being influenced by the introduction of artificial intelligence (AI) to the computer-aided drug design arena. In this regard, AlphaFold is an AI-based program that is used to generate the structure of macromolecules. It is possible to get the AlphaFold predicted structures of macromolecules from the UniProt database. Researchers might prefer to use the ready-made structures available in the UniProt. This study aimed to explore the reliability of the available AlphaFold structures relative to the structures determined through I-TASSER and SWISS-MODEL.
In this study, five drug targets with undetermined structures (human alpha-glucosidase, AXL, CEMIP, DYRK1B, and PARP9) were selected. The structures of the targets were generated through I-TASSER and SWISS-MODEL. The AlphaFold predicted structures of the targets were retrieved from the UniProt database. The best structure of the targets among the I-TASSER and SWISS-MODEL generated models was selected based on ERRAT, Verify3D, and Ramachandran plot criteria. The best structure was then compared to the AlphaFold predicted ones by using ERRAT, Verify3D, and Ramachandran plot values first. Thereafter, the comparison was pursued by using molecular dynamics (MD) simulation to evaluate the stability of the predicted structures.
This study revealed that the AlphaFold structures had comparable reliability to the other two methods in terms of ERRAT, Verify3D, and Ramachandran plot values. Furthermore, none of the programs were found to be superior in every aspect. However, the AlphaFold predicted structures had low stability in the MD simulation study. In short, the computational study emphasized the role of consensus modeling in generating reliable structures.
Keywords: drug design, drug target, homology modeling, MD simulation, structure validation
Audience Take Away Notes:
- The audience will be able to design three-dimensional (3D) structures of drug targets more accurately
- The knowledge from this presentation will simplify the structure-based drug design process for researchers
- Several programs and servers are used in homology modeling. Researchers might be confused about the right way towards a reliable structure modeling. The presentation will present the most suitable approach to get a reliable 3D model
