A computational workflow for binding free energies in Python

Nygaard-Thomsen, Simon

4. Term

Chemistry, Master (polyt)

2022

2022-09-08

37

β-cyclodextrin (β-CD) can trap small, poorly soluble molecules in water by forming inclusion complexes. This study tested whether widely used molecular dynamics tools can capture this behavior and quantify how strongly the guest binds. Using Python, we ran simulations in OpenMM with the OpenFF/SMIRNOFF force fields, which assign parameters by recognizing chemical patterns (SMIRKS) directly from the molecular graph—a more flexible approach than older force fields such as AMBER. To compute the binding free energy—a measure of the strength of association between β-CD and a ligand—we applied a biasing potential along their inter-molecular distance and performed Umbrella Sampling in overlapping windows. We analyzed the data with FastMBAR, a tool that applies the Bennet Acceptance Ratio, to estimate free energies and uncertainties. The results show that OpenMM is a viable option for this problem. With about 2000 snapshots per window and a bin size of 20 per window, the free energy estimates and their standard deviations stabilized. Comparison with literature indicates the method is feasible, although some predicted binding free energies were somewhat high.

β-cyclodextrin (β-CD) er kendt for at kunne indkapsle små, dårligt opløselige molekyler i vand og danne inklusionskomplekser. I dette projekt undersøgte vi, om udbredte simulationsværktøjer kan beskrive denne adfærd og samtidig kvantificere, hvor stærkt molekylerne binder. Med Python kørte vi molekylærdynamiske simuleringer i OpenMM sammen med OpenFF/SMIRNOFF-kraftfelter, som tildeler parametre ved at genkende kemiske mønstre (SMIRKS) direkte i molekylets graf—a en mere fleksibel tilgang end ældre kraftfelter som AMBER. For at beregne bindingsfri energi—et mål for, hvor stærkt β-CD og en ligand associerer—påførte vi et biaspotentiale langs afstanden mellem de to molekyler og udførte Umbrella Sampling i overlappende vinduer. Data blev analyseret med FastMBAR, et værktøj der anvender Bennet Acceptance Ratio, for at estimere frie energier og usikkerheder. Resultaterne viser, at OpenMM er et anvendeligt valg til denne opgave. Med cirka 2000 snapshots pr. vindue og en bin-størrelse på 20 pr. vindue stabiliserede de frie energier og standardafvigelserne sig. Sammenligning med litteraturen indikerer, at metoden er brugbar, om end nogle af de beregnede bindingsfrie energier var lidt høje.

[This apstract has been rewritten with the help of AI based on the project's original abstract]

bcd ; cyclodextrin ; force fields ; glycam06 ; SMIRNOFF ; OpenMM ; OpenFF ; inclusion complex ; ligand ; guest-host interaction ; PMF ; free energy ; FastMBAR ; MBAR ; BAR ; Bennet Acceptance Ratio ; computer chemistry ; simulations ; MD ; Molecular dynamic ; supra chemistry ; aspirin ; benzene ; phenol ; Sage ; openforcefield ; python

Download PDF
View record in AAU Student Projects