Skip to main content
2024 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

High School Student with a Traditionally Underrepresented Background

HS Poster

Poster 115 - Molecular Docking Analysis of Peptide-Protein Interactions for Consensus Scoring and Binding Affinity Prediction

Saturday, October 26, 2024
10:00 AM - 12:00 PM EST
Location: Exhibit Hall E, F & G

Introduction: Peptide-protein interactions (PepPIs) are highly abundant in living cells and play a pivotal role in a wide range of cellular processes, such as cell signaling, immune recognition, enzyme inhibition, gene expression regulation, etc. For this reason, peptides have become promising drug candidates, and more than 80 peptide drugs have been approved by FDA. However, peptides, for being shorter than a full protein, usually can assume very different conformations when in association with a protein. To distinguish between such different poses, a comprehensive understanding of PepPIs is vital for rational design of novel peptide therapeutics. For the same reason, it remains laborious, time-consuming, and costly to determine PepPIs experimentally. Due to this limitation, many tools have been developed to help predict such peptide poses. And, while many tools are available, they are usually specific to a certain group of proteins due to the difficulty of dealing with the particularities of each protein family. Machine learning methods, however, have changed this landscape and one new barrier is the availability of consistent experimental data for training such models. The publicly available experimental affinity datasets are still limited by different affinity measurement methodologies, inconsistent experimental conditions, and lack of 3D binding structural information. To mitigate this issue, we are working on creating a consistent dataset to be used as training input in a consensus scoring model based on the correlation of the scoring provided by MOE with the available experimentally measured affinity values (IC50, EC50, Kd, and Ki) of the selected PepPIs.

Materials and

Methods: In this study, the computational structure-based molecular docking and simulation software MOE (Molecular Operating Environment) [1] was used to calculate the scoring of peptide-protein pairs. First, the peptide-protein subset of data was obtained from Propedia database [2] that matches the IDs with the BioLip database [3]. These two databases comprise biologically relevant high-resolution structures from the Protein Data Bank including structural and sequence information from protein-peptide complexes. To score the receptor-ligand complexes, the Quick Prep panel on MOE was run to repair any missing atoms from the PDB files. Next, we open the template docking mode in the dock panel, where the receptor and ligands were already identified by MOE. If they were not, the receptor and ligand(s) were manually selected in their respective dropdowns. Once the PepPIs docking were done, MOE database file was generated with the confirmations and their properties, along with their scores. In total, 257 PepPIs were rescored and combined into a labeled dataset to be used as input on the Supervised Machine Learning training model.

Results, Conclusions, and Discussions: The output 3D molecular structure from the PepPIs were saved and the scores (S) ware annotated. From the BioLip database we extracted the experimental information and the score of each rescored crystal was correlated with the experimental information. The ligand root mean square deviation (RMSD) values were also computed and ranked as part of the correlation system for the PepPIs. The resulting data will be used as part of the input on the Supervised Machine Learning training model to build a consensus scoring.
In conclusion, computational docking allows modeling the 3D structure of peptide-protein interactions, providing structural insights on key molecular determinants of PepPIs. The consensus scoring model can serve as a useful tool in peptide-protein interaction prediction and identification of important binding residues in the peptides, which can thus facilitate the peptide drug discovery process.

Acknowledgements (Optional): This work was funded by a UH Drug Discovery Institute Seed Grant.

References (Optional): [1] Molecular Operating Environment (MOE), 2022.02 Chemical Computing Group ULC, 910-1010 Sherbrooke St. W., Montreal, QC H3A 2R7, 2024.
[2] Martins PM, Santos LH, Mariano D, Queiroz FC, Bastos LL, Gomes IS, Fischer PHC, Rocha REO, Silveira SA, de Lima LHF, de Magalhães MTQ, Oliveira MGA, de Melo-Minardi RC. Propedia: a database for protein-peptide identification based on a hybrid clustering algorithm. BMC Bioinformatics. 2021 Jan 2;22(1):1. doi: 10.1186/s12859-020-03881-z.
[3] Yang J, Roy A, Zhang Y. BioLiP: a semi-manually curated database for biologically relevant ligand-protein interactions. Nucleic Acids Res. 2013 Jan;41(Database issue):D1096-103. doi: 10.1093/nar/gks966. Epub 2012 Oct 18.