TR&D1:  Molecular modeling and simulations: Bridging molecular and cellular scales

Vision. There is a growing need to understand molecular events at the mesoscopic time scale - microseconds-to-seconds, for systems containing 10s-to-100s of proteins/subunits, which current methods usually fail to represent with adequate structural and spatial complexity. We also have new challenges with 'omics'-scale data, which could be best tackled by advanced algorithms and high performance computing (HPC) resources. Significant progress has been made in the TR&D1 project during the past funding period, evidenced by 38 publications that acknowledged the P41 support, made by TR&D1 members.1-38 We developed and disseminated novel computational technology, and helped accelerate biomedical research driven by two DBPs. Many tools that we developed in the past decade, rooted in fundamental concepts of statistical mechanics, spectral graph theoretical methods and machine learning, can now be substantively advanced to meet the emerging needs and challenges. TR&D1's overarching goal is to develop, implement, integrate and apply computational technology toward meeting the emerging needs for structure-based modeling of mesoscopic- and/or omics-scale dynamics, and to establish a platform that synergistically interfaces with the technologies developed in the other TR&Ds

timescales

Time scales sampled by molecular (MD) and subcellular (MCell) simulations. The intermediate regime, mesoscale, is poorly sampled. Elastic network models aim at filling the gap between those scales.

Background and Motivation. Biomolecular systems have access to multiple substates under physiological conditions. Examples are the open and closed substates of enzymes around their substrate-binding site (inter-domain cleft), the inward-facing or outward-facing substates of transport proteins, or the substates visited by molecular machines during their allosteric cycles. It is now widely established that the passage between these functional substates is facilitated, if not intrinsically favored, by the overall architecture/fold of the proteins. This property, presumably optimized by evolutionary selection of functional folds or assemblies, is usually robust to atomic or energetic details, such that coarse-grained (CG) analytical approaches permit us to gain insights for the first time, albeit at low resolution, into the substates and mechanisms accessible to multimeric structures. Recent years have seen an explosion in the number of such CG approaches, including in particular those based on elastic network models for complex structures. These approaches may potentially start to fill the gap between those adopted for modeling molecular dynamics and microphysiological events.

In parallel with our improved understanding of the role of structural dynamics in defining biologically relevant substates and mechanisms, there have been significant advances in computing technology in recent years, which now allow us to examine larger systems, longer time scales than those accessible by classical molecular dynamics (MD) simulations. We now generate microseconds-to-milliseconds-long MD trajectories (depending on the size of the examined system) and it becomes imperative to develop and implement automated analytical methods for efficient extraction of useful properties from those detailed trajectories. These properties may, in turn, be utilized as input for lower-resolution simulations, and so on, thus opening the way to develop methodologies that can explore processes of biological interest at hierarchical levels of resolution.

 

Goals

The overarching goal of the TR&D1 is to take advantage of advances made in both biomolecular modeling and computing technology toward developing and implementing computational tools for qualitative and quantitative elucidation of time-resolved molecular events that control neurobiological systems dynamics.  

In TR&D1, the focus is on the development of models, methods, and software for

  • predicting the structure-based dynamics and functional transitions of selected biomolecular systems at multiple scales
  • efficient evaluation of the thermodynamics and kinetics of biomolecular interactions
  • integrating the above models and methods across time and length scales by suitable APIs
  • generating models and methods portable to TR&D2 where possible