Tackling sampling and accuracy issues in biomolecular simulations

Computational modeling of a complex heterogeneous system presents two major challenges. First, the accuracy of biomolecular simulations based on physical models of the system is bounded by the limitations of the underlying theoretical description of the system. To improve the accuracy, one can properly integrate all sources of information available, such as physico-chemical and statistical a priori information as well as experimental measurements. Second, the efficient generation of models requires advanced sampling techniques to overcome the time scale limitations of traditional methods such as Molecular Dynamics or Monte Carlo. In these lectures, I will presents computational methods aimed at solving these two issues. In particular, I will introduce "Metadynamics Metainference", a combined approach that can tackle both problems at the same time. Metainference is a Bayesian inference method to optimally integrate the available information by dealing with all sources of noise and with the ensemble-averaged nature of experimental measurements. Parallel Bias metadynamics is a new metadynamics flavour that overcomes some of the limitations of the original metadynamics approach to enhance sampling in Molecular Dynamics simulations. I will present the theory of Metadynamics Metainference and its application to model problems in structural biology, including modeling of NMR and cryo-electron microscopy data.

Dr. Massimiliano Bonomi received his PhD in February 2010 in the group of Prof Michele Parrinello at ETH Zurich. His PhD studies concerned the development and application of enhanced sampling techniques to study rare events in biological systems. He is a creator and a core developer of the PLUMED project (www.plumed.org). After a postdoctoral experience in the group of Andrej Sali (University of California San Francisco), he currently holds a postdoctoral appointment at University of Cambridge in the Vendruscolo group, where he works on the development and application of Bayesian methods to incorporate noisy experimental data collected on heterogeneous samples into molecular dynamics simulations.