Free energy and hidden barriers: the β-sheet structure of the prion protein

SERGIO ALEXIS PAZ; CAMERON F. ABRAMS

San Diego

Congreso; 251st American Chemical Society National Meeting and Exposition.; 2016

Despite molecular dynamics (MD) being a powerful simulation tool, it becomes unusable for the prediction of the many common physico-chemical properties which require extremely long computational times. Thus, the development of new methods that can extend the temporal scope of MD is a major goal in computational chemistry. Following this direction, we introduced on-the-fly free-energy parameterization (OTFP) [1] as a new biasing free-energy method akin to metadynamics [2] with one important distinction: rather than using an history-dependent bias potential to enhance sampling, acceleration is achieved using temperature-accelerated molecular dynamics [3].Recently, we used OTFP to reconstruct the free energy surface (FES) of the β-sheet structure of the mouse prion protein [4]. Misfolding and aggregation of this protein is associated with a collective ofneurodegenerative diseases [5]. The stability of the small antiparallel β strands of the native conformation has been focus of attention in the attempts to understand the mechanism of these events [6]. The FESobtained via OTFP is found to be in good agreement with the NMR structure of the protein [7] but stands in notable contrast with previous results obtained via metadynamics [8]. Furthermore, near the β-sheet region we unveil the presence of hidden barriers that introduce a much higher degree of statistical noise in well-tempered metadynamics (WTMD) simulations than in OTFP. This sensitivity is shown to be aconsequence of the history-dependence bias potential of WTMD. In contrast, OTFP does not retain any memory of previous sampling in the bias potential and thus displays an improved convergence behavior.Slow hidden variables are a very common phenomena in attempts to recover free energy surfaces for new and complex systems. While OTFP appears to be a robust and preferable method to confront this issue, it is only able to enhance the sampling of the chosen collective variable. The combination of OTFP and replica exchange molecular dynamics (REMD) may be a solution to tackle both issues at the same time. Our preliminary tests for the new OTFP/REMD method had show very promising results.