(AHeDD2025) Quantitative Evaluation of Protein-Ligand Substructure Interaction with Inverse Mixed-Solvent Molecular Dynamics Simulation

Transcript

1. Asia Hub for e-Drug Discovery Symposium 2025 I Hangzhou, CHINA Quantitative Evaluation of Protein-Ligand Substructure Interaction with Inverse Mixed-Solvent Molecular Dynamics Simulation Keisuke Yanagisawa1, Ryunosuke Yoshino2, Genki Kudo2, Takatsugu Hirokawa2 1 Institute of Science Tokyo, Tokyo, Japan 2 University of Tsukuba, Ibaraki, Japan E-mail of the corresponding author(s): t-hirokawa@md.tsukuba.ac.jp This work was partly supported by JSPS KAKENHI (23K24939, 23K28185, 25K03215), Kayamori Foundation of Informational Science and Advancement, and AMED BINDS (JP25ama121026, JP25ama121029). Acknowledgements [1] Yanagisawa, K., Yoshino, R., Kudo, G., Hirokawa, T. IJMS 23: 4749, 2022. [2] Friberg, A., et al. JMC 56: 15-30, 2013. [3] Chessari, G., et al. JMC 58: 6574-6588, 2015. References In this study, we proposed a method to predict binding affinity with inverse MSMD simulation. Our method enables the affinity prediction without requiring additional simulations for target proteins, resulting in reducing computational cost. As a result, the corr. coef. reached r = 0.50 for the SAR dataset. We also showed the robustness to side-chain flexibility. Conclusion https://github.com/keisuke-yanagisawa/exprorer_msmd source code: MD engine GROMACS 2019.4 Force field Amber ff14SB (protein) GAFF2 (probes) Water model TIP3P Pseudo repulsion between probes LJ (σ = 2nm, ε = 1 × 10-6 kcal) between centers of probes Conc. of probes ≈ 0.25M Time step 2 fs Production run 800 ns (40 ns × 20 runs) per a probe Trajectory Snapshots every 10 ps, in 20-40 ns MSMD simulations with 15 diverse proteins were conducted per probe Introduction: Inverse MSMD Methods: Profile-Structure Matching Results and Discussion N N N OH N N Mixed-solvent MD (MSMD) MD in explicit water mixed with probe molecules Applications: Hotspot detection, Binding affinity prediction MSMD simulation Probe existence probability Probes Probe poses in MSMD snapshots Extract & aggregate Cβ atom positions Interaction profile (19 amino acids w/o Gly) Inverse MSMD [1] Estimating interaction profile of probes via MSMD Interaction profile: spatial Cβ atom existence probability Obtaining protein-independent interaction information MSMD simulation for a protein of interest is always required The profile might be applied analyses without additional MSMD Case 1: Binding affinity prediction (MCL-1 [2]) Correlation coefficient: 𝒓𝒓 = 𝟎𝟎. 𝟓𝟓𝟓𝟓 Case 2: Robustness evaluation of the method (XIAP [3]) The method is robust to side chain-level flexibility Initial complex (PDBID: 5C7A) N +H2 N O N N Cl N Crashed to Lys substitution point score = 29.79 (IC50 = 𝟎𝟎. 𝟔𝟔𝟔𝟔 μM) score = 37.86 (IC50 = 𝟎𝟎. 𝟏𝟏𝟏𝟏 μM) < Probe superimposition 4 5 6 7 8 40 60 80 100 120 −log(𝐾𝐾𝑖𝑖 ) sum of matching scores substitution 1 substitution 2 Initial complex (PDBID: 4HW3) N O O - S Cl O O - HN Cl O - O O Cl O O - … F F F O Cl … Step 1: Superimpose the profile to an initial structure Step 2: Obtain pseudo energy score of Cβ positions pseudo energy score = ln(relative Cβ occupancy) Step 3: Calculate matching score by summing up scores of all Cβ positions in the protein 0.7 0.2 0.4 Initial structure Interaction profile (spatial Cβ occupancy) = ln Cβ occupancy in profile bulk occupancy MSMD protocol Results QR code for this poster