Diffusion coefficients are routinely estimated from molecular dynamics simulations by “fitting a straight line” to observe mean-squared displacements (MSDs). Typically, this fitting is performed without considering fundamental concepts of displacements, such as their heteroscedasticity and correlation, leading to a sub-optimal estimation method. These diffusion coefficients are essential in modelling temperature-dependent behaviour and determining diffusion’s activation energy and mechanism.

I will present a Bayesian scheme for estimating the diffusion coefficient from a single simulation trajectory with high statistical efficiency and accurately estimating the uncertainty in the predicted value [1]. I will then discuss the impact of accurate uncertainty estimation on downstream analysis, including temperature-dependent behaviour [2, 3]. Throughout this seminar, I will introduce kinisi, a software package developed to improve the analysis of molecular dynamics simulation [4].

You may enjoy this seminar if you are interested in data analysis, computational modelling, or open-source software.

1. A. R. McCluskey, S. W. Coles and B. J. Morgan, arXiv:2305.18244v5, 2024. 2. A. R. McCluskey, J. Chem. Educ., 2023, 100(11), 4174–4176. 3. A. R. McCluskey, S. W. Coles and B. J. Morgan, In Preparation, 2024. 4. A. R. McCluskey, A. G. Squires, J. Dunn, S. W. Coles and B. J. Morgan, Journal of Open Source Software, 2024, 9, 5984.