Abstract:Contemporary structural biology has an increased emphasis on high-throughput methods. Biomolecular simulations can add value to structural biology via the provision of dynamic information. However, at present there are no agreed measures for the quality of biomolecular simulation data. In this Letter, we suggest suitable measures for the quality assurance of molecular dynamics simulations of biomolecules. These measures are designed to be simple, fast, and general. Reporting of these measures in simulation papers should become an expected practice, analogous to the reporting of comparable quality measures in protein crystallography. We wish to solicit views and suggestions from the simulation community on methods to obtain reliability measures from molecular-dynamics trajectories. In a database which provides access to previously obtained simulationssfor example BioSimGrid (http://www.biosimgrid.org/)sthe user needs to be confident that the simulation trajectory is suitable for further investigation. This can be provided by the simulation quality measures which a user would examine prior to more extensive analyses.
OVERVIEWFor the past quarter century, biomolecular simulations have been adding value to structural biology via the provision of dynamic information. 1 As genomics move from sequencing to structural and dynamical considerations, and highthroughput technologies advance from crystallography to molecular-dynamics (MD) simulation, this process is occurring with vigor. As the bibliometric data in Figure 1 show, MD simulation of biopolymers is now becoming a routine technique. To help this maturation process, standardized practice should be established in the simulation community, similar to that in crystallography.2,3 It is already regular practice to print quality measures in a formulaic table in published articles reporting crystallographic resultssindeed, it is surprising if such a table is missing, and the referees would readily reject the manuscript.We are hereby initiating a discussion on the appropriate measures of quality and convergence 4 for MD simulation trajectories of biopolymers. The process of calculating these measures is designed to be automated for large numbers of trajectories; hence the set of analyses used for this description should be general, with minimal interaction of a human curator. The scientist can then use these measures, along with sensible comparisons with known experimental data (which we recognize as essential), to decide whether a specific trajectory is suitable for further investigation. Our purpose is to solicit feedback from the simulation community with regard to the analyses we have chosen and to obtain further suggestions. We invite the community to express their views on our choices of measures.We are motivated to do this by our work in building BioSimGrid, 5 a distributed environment for archiving and analyzing biopolymer simulations. Other similar databases are emerging 6 (personal communications with Valerie Daggett and Modesto Orozco, http://mmb.pcb.u...