Revealing pathways and kinetics of molecular recognition with advanced molecular simulation algorithms

点击翻译按钮获取中文摘要

Project Summary Biology at the nanoscale is driven by molecular recognition. Cells are lined with receptors that recognize a myriad of biomolecules and transmit signals to the cell interior. This forms the basis of cell-cell communication that allows multicellular organisms to thrive. An understanding of molecular recognition is tremendously important for human health, as it underlies our immune system’s response to threats from viruses, bacteria and cancer cells. Unfortunately, these recognition processes are often much more complicated than the canonical “lock-and-key” paradigm. Many neuropeptides and peptide hormones are dozens of residues in length and can exhibit tremendous structural plasticity. The study of molecular recognition in such dynamic systems is a challenge, as only a limited understanding of the interactions can be gleaned from structural approaches alone. Computational molecular dynamics (MD) simulation allows us to view complex biomolecular systems in atomic detail, allowing us to compute binding and unbinding rates, and to study their molecular determinants. A well-known drawback of MD is the difficulty of exploring conformational landscapes without getting trapped in local free energy minima. The PI (Dickson) has over a decade of experience developing new methods for tackling this problem, which are able to calculate transition rates in complex systems. Particularly, the REVO method (“Reweighting Ensembles by Variation Optimization”) has shown success on unbinding pathways of drug-like ligands with mean first passage times up to 34 minutes, which is billions of times beyond the typical reach of straightforward MD. This is done without applying biasing forces to the system or making any assumptions about equilibrium conditions. REVO is essentially an evolutionary algorithm in trajectory space, where a large group of trajectories are run in parallel, and outlier trajectories are identified using a measurement of distance to other ensemble members. These outliers are “cloned”, which increases the probability of seeing long-timescale events happen, even in short-time trajectories. The proposed work will augment this method with recent advances in the automatic detection of slow collective variables (or “CVs”). The VAMPnet (“Variational Approach for Markov Processes”) approach will be used to detect groups of CVs to use in the REVO distance calculation. A method for iteratively improving the accuracy of rate calculations (“Asynchronous weighted ensemble”) is also proposed. Together these developments will constitute a powerful tool for studying long-timescale events in complex biomolecular systems that will be made freely available to other researchers. Through collaborations with experimental partners, this method will be applied to reveal the molecular mechanisms of: i) peptide signaling for a class-B GPCR (Aim 2), and ii) specific recognition of bacterial LPS by a phage tail-spike protein (Aim 3). These mechanisms will inform the design of new molecules with potential therapeutic applications.