Bristol Centre for Complexity Sciences Exploring methods for analysing trajectories of long-scale Molecular Dynamics simulations on the example of the 150-loop in Neuraminidase

The development of ever more efficient hardware and software has allowed for the simulation of complex biological systems, such as that of a large protein, in atomistic detail on the timescale of up to around a microsecond. This is enough to observe biologically significant conformational changes driven by complex combinations of thermally induced perturbations to single atoms that make up the protein. Neuraminidase (NA) is a major protein of influenza (commonly known as flu), the causative agent of many recent pandemics. Recent studies have revealed complex functional dynamics of the protein’s 150-loop, which leads to drug unbinding and confers drug resistance. This project involved running simulations of 7 types of NA (including wild-type and mutated structures of the free protein, as well as bound with Oseltamivir) in order to find out what underpins the mechanism of drug resistance and explore data mining and machine learning methods for identifying biologically significant events in the simulations and quantitatively defining the conformational states of the 150-loop. The simulations confirmed previously observed behaviour of the 150-loop that can be mechanistically related to drug resistance. Reducing the dimensionality of the system down to only the states that are defined by distinct peaks in the time occupancy of the secondary structure dihedral angle histograms and constructing a network of transitions between states provides a new method for comparing the phase space explored by simulation trajectories and quantifying activation energy barriers between conformational states.