A computational workflow for binding free energies in Python

Student thesis: Master Thesis and HD Thesis

  • Simon Nygaard-Thomsen
4. Term, Chemistry, Master (polyt) (Master Programme)
The literature show that $\beta$-cyclodextrin ($\beta$-CD) have good probabilities of forming inclusion complexes in water with ligands having a low solubility themselves, both \textit{in situ} and \textit{in silico}.
Through Python, the framework OpenMM was used together with OpenFF (developers of the SMIRNOFF force fields (FF)) to simulate such inclusion complexes in water with Molecular Dynamics (MD). SMIRNOFF uses so-called direct chemical perception, in opposition to older FF like AMBER, to parameterise molecules directly from the chemical graph by the use of SMIRKS. SMIRKS are able to recognise patterns in molecules, making the process more efficient.
The goal of the MD simulations was to introduce a biasing potential between the $\beta$-CD and a ligand and by measuring the distance between the two molecules in each snapshot, finding the free binding energy of the inclusion complex.
Through Umbrella Sampling (US) and analyses by FastMBAR, a software tool for applying the Bennet Acceptance Ratio, it was shown that OpenMM was indeed a viable option in this regard.
Through the analyses, it was revealed that a sample size of 2000 snapshots per window in the US and a bin size of 20 per window, both the free energy and standard deviations converged. Comparisons with the literature showed that the method was feasible although some of the free energies of the inclusion complexes were a bit high.
Publication date8 Sept 2022
Number of pages37
ID: 485455362