Proteins in realistic environments: simulation meets experiment
Please visit our instruction site for additional information
Proteins in a bioreactor or in a cell are solvated in a complex mixture of highly concentrated components which differs considerably from the dilute solutions of most in vitro and in silico studies. In realistic environment the kinetics of an enzymatic reaction does not only depend on the chemical structure of the substrate and the enzyme, but is described by a multidimensional parameter space: concentrations of the enzyme and additional proteins, state of the enzyme (monomeric, multimeric, clusters, or immobilized), protein sequence, concentrations of substrate(s) and of co-substrate(s), concentrations of product(s) and side-product(s), components of the solvent (water, organic solvent, additives, salts), temperature, and pressure [Ga10]. It is a major challenge for experimentalists and for modelers to explore this huge parameter space, and strategies such as high-throughput experimentation, workflow system for parallel modeling, and statistical analysis or deep learning methods to reduce the dimensionality of parameter space, are highly desirable.
Deterministic non-steady-state kinetic models are classically used in the analysis of enzyme dynamics by quantifying the individual binding and dissociation steps of reactants and also the catalytic steps that are obtained e.g. by stopped or quenched-flow or spectroscopic methods. These non-steady-state kinetic models allow evaluating the effect even of individual amino acids on the enzyme function with respect to substrate orientation, active site environment, transition states, and protein dynamics [He13]. In contrast, steady-state mechanistic models mostly build on the experimental analysis of dynamic reactant concentrations, either as initial rates or progress curves. However, the identification of macroscopic kinetic parameters from real experimental data is by far not straightforward, but requires appropriate workflows which are supported by a suitable experimental design [Al12]. An approach which combines experimental techniques with modelling methods on macroscopic and microscopic scales is expected to provide mechanistic insights into the complex interactions that mediate biochemical and biophysical properties of proteins in realistic environments [Sa15].
On the macroscopic scale it has been demonstrated how the application of thermodynamic principles to enzyme-catalysed reactions, e.g. by equation-of-state approaches, are a powerful tool to assess biological reactions and evaluate the influence of reaction conditions on the reaction equilibrium [He16].
On the atomistic level, free-energy simulation methods based on statistical mechanics and molecular simulations are considered to be the most rigorous route to predict a wide range of thermodynamic properties in various research areas including solvation thermodynamics, molecular recognition and protein folding. In contrast to equilibrium binding affinity much less is known about the molecular determinants of binding kinetics because the rate constants inherently depend on the free-energy barriers. Although microsecond timescale association events are becoming accessible for atomistic simulations nowadays, brute force equilibrium simulations will be prohibitively expensive for ligands of submicromolar affinities. For such cases, the much slower dissociation rates still require novel algorithms to bridge the timescale gap which are usually based on path sampling approaches that aim to capture rare events by minimizing the simulation of long waiting times between events [Pa17,El17]. Moreover, most MD studies largely focus on single molecules or complexes without considering crowded environments whereas the consideration of realistic concentrations may provide new important insights [Fe17].
A quantum mechanical description of the interatomic forces is typically required for the chemical step during an enzymatic process, while binding and unbinding and possible structural rearrangements can often be described by force fields with sufficient accuracy. QM/MM with a density functional theory or higher description of the QM part is usually computationally prohibitively expensive for full sampling of free-energy differences. However, the main entropic changes during the chemical step of an enzymatic reaction are often caused by the environment of the active centre, which is described by a force field. The QM part is typically responsible for the main enthalpic changes. QM/MM combined with free-energy perturbation offers a possibility to sample the MM part while keeping the QM part frozen. The method has been shown to provide reliable results for enzymatic reactions [Cu17].
New questions in and around enzymology lead to new types of experimental data sets that challenge the application of the macroscopic models:
• The particular restriction of the steady-state assumption may be violated in particular when combining individual enzymes to signaling networks or to metabolic networks.
• High throughput experimentation may provide the amount of experimental data necessary in order to cover the mentioned experimental parameter space (concentrations, pH, temperature,…), but the corresponding ‘big’ data analysis poses new challenges.
• When extending the experimental parameter space to the sequence space, like in enzyme design or enzyme evolution, computational methods have to support the interpretation of changes to the kinetic parameters or even the reaction path with respect to structure-function relationships.
• Single-molecule measurements introduce stochastic data to enzymology, which allows analyzing the minimal number of molecules required for deterministic models.
These new challenges render the integration of macroscopic models with free energy simulations or QM/MM methods highly desirable. The goals of the workshop are therefore to improve interaction and mutual understanding between the different worlds of modeling and experimentation, to provide an inspiring environment for creating innovative ideas and establishing new collaborations, and to push forward the molecular interpretation of experimentally obtained kinetic data. Thus, the workshop will serve to bridge between different scales and between scientific fields thereby focusing on the particular application of enzymatic reaction in realistic environments. In particular the following points will be discussed:
• How to quantitatively describe fundamental thermodynamic properties such as the chemical potential in crowded solutions?
• How can approaches on different scales, i.e. equations of state vs. atomistic simulation vs. ab initio methods benefit from each other?
• How does macromolecular crowding influences the thermodynamics and kinetics of biological reactions?
• How can simulations and experiments be coordinated to get insights into proteins in realistic environments?
• When the availability and power of computing rapidly increases would it be feasible to apply more complex force fields in atomistic simulations?
• Are machine learning approaches able to optimize transferable parameters for the underlying force field functions or may they even replace them?
• How to handle sparsity and variability of experimental data for dynamic modeling?
We hope to work out how scientific knowledge established on a particular scale of modeling can be beneficial for model building and/or model validation on another scale, be it through defining boundaries for particular parameters, elucidation of reaction mechanisms, new tools for analysing experimental data etc. This should work in both directions, bottom-up and top-down.
Niels Hansen (Institute of Thermodynamics and Thermal Process Engineering, University of Stuttgart) - Organiser
Jürgen Pleiss (Institute for Technical Biochemistry, University of Stuttgart) - Organiser
Antje Spieß (Technical University of Braunschweig) - Organiser