Fast Energy Evaluation for Multi-Protein Systems

Simulation of large multi-component molecular systems with atomistic details is an important step toward understanding cellular processes. In this proposal we check the feasibility of addressing the common bottleneck in this class of difficult problems, namely energy evaluation. The basis for our approach is the observation that the interaction energy between two proteins (or discrete protein conformations within an ensemble) – can be efficiently calculated over the entire rotational-translational space using the fast Manifold Fourier transform (FMFT) correlation approach. Given any conformation of the complex multi-particle system, its energy can be easily obtained by summing the pairwise interaction energies extracted from the lookup tables. The key innovation to efficiently implement this method will be our ability to compress and store the interaction energy lookup tables in memory using wavelet sets, similar to the ones used for image compression in the JPEG 2000 algorithm. In addition compressed table approach will allow us to account for side chain dynamics. The use of this innovative approach will speed up energy evaluation by at least several order of magnitude, with reasonably accurate atomistic potentials, and will open new opportunities, including detailed exploration of the conformational space by Monte-Carlo like simulations in systems that were considered to be too large for proper sampling with traditional energy evaluation. We will apply the methodology for simulations of multi-protein assemblies in conjunction with low resolution Mass Spectrometry (MS) data, providing mechanistic insight into cellular function . .