Background and significance
Ever increasing advances in structural biology are challenging the field of multiscale modeling with the need of going beyond the limits of time scale and system-size of molecular simulations, modeling biological systems from the atomistic to the cellular level (CurrOpinStructBiol 2017, 43, 1). Thus, extending the simulations capabilities beyond the atomistic representation, invading the domains of bioengineering and bioinformatics, is of increasing importance for pursuing effective integrative research.
The aim of this workshop is to discuss these challenges, from atomistic to coarse grained (CG) and mesoscale (MS) representation of biomolecules into the cell environment, exploring novel integrative methods from the macromolecular (nm) to sub-cellular (μm) levels (AccChemRes 2017, 50, 594). These will include different kinds of phase space sampling, such as Brownian Dynamics (BD), advanced sampling methods and integrative methods that allow building structurally detailed 3D models of supra-molecular structures at cellular level.
Clearly, the limits of time scale and system-size of simulations are a bottleneck when computations need to match experiments. With respect to the time scale, enhanced sampling methodologies, such as accelerated Molecular Dynamics (MD), allow routine access to millisecond events. However, for big-size systems, the configurational space sampling often results poorly explored, resulting in a large statistical noise, which makes difficult the derivation of well-converged free energy profiles.
With respect to the system-size, CG and MS methods enable the reduction of the degrees of resolution, reaching the (sub)cellular level by “coarse graining” at different resolution (AccChemRes 2010, 43, 220). CG leaves to the user the definition of the interacting centers, opening issues regarding the models parameterization. MS models extremely coarse the system with the need to ensure accuracy re-including the relevant internal degrees of freedom. As such, MS (and CG) models are being integrated with non-particle approaches (elastic networks, continuum membrane models), invading the domain of bioengineering. This allows reaching the μm domain with relatively modest computational resources and opens the road to integrative approaches including system biology and bioinformatics, linking molecular modeling and bioengineering. The reliability of these methods with respect to cryo-EM and electron tomography is a main open challenge, which is being addressed by considering multiple strategies, with inclusion of the experimental information via machine learning approaches or by exploring the vast conformational space available to individual interacting proteins using Monte Carlo or MD techniques (AnnuRevBiophys 2016, 5, 253). Finally, integrative methods revealing the architecture of large molecular complexes and cellular portions (e.g., IMP, Haddock, CellPACK) are going to be interfaced with atomic-level codes. Following the philosophy of “multiscaling”, by re-including the atomistic representation into the supra-molecular description, this is thought to drive one of the next-years challenges in biophysics, which is the simulation of the cellular components.
Overall, our workshop will touch all the major unresolved aspects that today represent the challenges in multiscale modeling, fostering discussion between leaders in the field and young scientists, with the aim of advancing the field and develop new computational tools for the future of biophysics.
Our workshop will revolve on three specific goals, as detailed below.
1. Towards longer time scales.
This section is dedicated to the time scale problem. Although enhanced sampling methods and BD allow sampling configurations over milliseconds and seconds, much has to be done such that computational data could match time-resolved experiments (FRET, NMR). We will focus on the need of developing rigorous reweighting algorithms, usually suffering from large statistical noise, for obtaining meaningful equilibrium data and energy landscapes from enhanced sampling simulations. We will also discuss the current challenges in simulating long-time scale processes, such as diffusion and macromolecular associations, via BD. While novel BD implementations have shown to overcome the rigid-body approximation, user-defined parameters controlling solvent dielectric, hydrodynamics, desolvation, and ion screening can affect the realism of the solvent model and the consistency of the result with respect to kinetic data.
2. Extending the system-size beyond the atomistic representations.
This section focuses on extending the system-size by using methods beyond atomistic representations, such as CG and MS, for higher or lower resolution, respectively. These methods, which enable the description of supra-molecular structures (viruses, organelles, synaptic vesicles), will be discussed in light of their combination with non-particle methods such as continuum membrane, which is a frontier topic. Issues arise from the inclusion of the membrane elasticity and to its interaction with the particle-like representation of the cytoplasm, as well as when diffusion of membrane proteins has to be considered. Additional issues arise from the development of methods enabling the construction of large-scale models of membranes including curvatures. These problems are under debate and will be discussed in this workshop.
3. Multiscale modeling toward the simulation of the cell.
Here, we will discuss the integration of MS and bioengineering methods to realize large-scale models of the cell. Models of large biomolecular complex and cellular components obtained by CG-MS modeling supported by Cryo-EM data will be combined by means of integrative algorithms, incorporating data from systems biology and structural biology to provide models of (sub)cellular architectures. The reliability of these methods with respect to cryo-EM and electron tomography is a main open challenge, which is to be addressed by considering as a linear combination the physico-chemical properties of the system (with particular attention to inter-macromolecular binding affinity), or by using machine learning approaches on a large amount of heterogeneous data. Ongoing efforts are in extending the capabilities of these algorithms to create full-scale atomic-level models that can be used as input for MD simulations. Attention will be given to these open challenges, while also disussing the integration of these algorithms with the BD and with the Adaptive Poisson Boltzmann Model (APBS) software, for enabling to compute diffusion processes at the cellular level.