Mounting evidence from numerous experimental [1, 2] and computational studies [3] has demonstrated that biomolecules have motions that span a wide range of time and spatial scales. Some of those motions reflect the importance of maintaining a “minimal” level of flexibility for function. For example, an insightful analysis [4] based on the Lindemann criterion indicated that the surface of proteins is liquid-like while the core is solidlike. This result makes intuitive sense in that the solid core is important for stability while the fluidic surface is essential for the structural changes required for functional aspects such as ligand binding. In addition to such “generic” thermal fluctuations, it is generally agreed that there are also “functional motions”, which have specific characters (in direction, magnitude and time-scale) that make these motions essential to the unique function of a particular biomolecule. These range from structural transitions at the domain scale, which are implicated in the function of many “biomolecular machines” [5] and multi-subunit enzymes, [6] to relatively localized vibrations that have been proposed to facilitate chemical reactions. [7] Despite their biological importance, functional motions are difficult to identify and characterize at a quantitative level. One primary difficulty is that although experiments have progressed to the extent that single vibrational periods can be resolved in time, associating this with specific motions at the atomistic level is typically not possible. Therefore, a combined computational/experimental approach is necessary for understanding functional motions at atomic resolution. In other words, with the development of novel experimental techniques, new theoretical and computational methods need to be developed to help interpret new experimental observables. This joint computational/theoretical-experimental effort is required to better characterize and understand the working mechanism of functional motions in biomolecules. This in turn requires that computational and experimental research needs to be enmeshed much more closely and each side understands the potential and limitations associated with specific experiments or simulation techniques. On the practical side, we note that dramatic progress has been made in computational design of enzymes [8]. To make the design more effective, especially for a high level of catalytic efficiency, an important challenge that needs to be tackled is the consideration of protein motion. Therefore, understanding the mechanism of functional motions has not only fundamental value but also direct impacts on our ability to design biomolecules of novel functionalities.

As pointed out in Objectives, the main goal of the workshop is to stimulate interactions between leading experimentalists and theorists to discuss problems, in the context of functional motions in biomolecules, for which new theoretical models and computational methods are needed. Therefore, we focus below on the key state-of-the-art theoretical methods. On the experimental front, novel spectroscopic techniques for studying (bio)molecular motions include multi-dimensional infrared spectroscopy [9], new nuclear magnetic resonance relaxation techniques [10] and single molecule fluorescence spectroscopy [11]. We plan to invite representative researchers who are active developers in these areas (Hamm, Martin, Bruschweiler, Kay, Wand, Yang, Noji and Zhao).

For studying a reactive event directly, a potential function that allows bond formation and breaking is required, which can be based on either the combined Quantum Mechanical/Molecular Mechanical (QM/MM) approach or a reactive force field such as the Empirical Valence Bond (EVB) formulation. QM/MM methods are more general and many successful methods are available [12], although practical studies have to carefully balance computational speed and accuracy; reactive force fields remain a competitive option for studying dynamics effects especially when the type of reaction is known [13]. For studying long-range electron transfer dynamics, for example, novel QM/MM methods need to be developed to allow the inclusion of protein/solvent dynamics [14, 15, 16]. For a quantitative description of multi-dimensional infrared spectroscopy, novel effective potential functions are also needed to describe how vibrational properties of the chromophore depend on the environment [17, 18].

For studying the dynamics of relatively fast reactive events, including many electron transfers, appropriate techniques are activated dynamics and transition path sampling (TPS). The former starts molecular dynamics (MD) trajectories from transition states (TS) and therefore requires identification of the (approximate) TS first by, for example, umbrella sampling calculations along an assumed reaction coordinate. The TPS technique [19], by contrast, searches barrier crossing trajectories directly without any presumed reaction coordinate and transition state, although the computational cost is substantially higher than activated dynamics simulations. Both methods make possible the analysis of the nature and time scale of protein motions that actively participate in the barrier-crossing process for the chemical step, although TPS is less biased by the user input. Extending the applicability of TPS to complex systems is an active area of research in modern molecular simulations [20]. For studying the slower conformational dynamics that may control (gate) the chemical event, free energy (potential of mean force) simulations along relevant conformational coordinates are useful for dissecting the energetics and kinetic bottleneck of the relevant motions. However, it is often true that the motions are complicated and can not be adequately described by a simple reaction coordinate. Therefore, an active topic of research involves methods that either directly map the minimum free energy path [21] or construct an effective low-dimensional representation (e.g., diffusion map) [22] for the motion of interest. Alternatively, one can construct network models for the complex conformational dynamics if the stable basins and transition rates among them can be robustly constructed based on multiple simulations [23]. Such approaches can also be successfully applied to ligand migration in proteins. [24] Finally, due to the often collective nature of slow conformational motions, useful insights can often be gleaned through coarse-grained (CG) models [25], which reduce the resolution of either temporal or spatial representation; example for the former is the coarse-molecular-dynamics approach of Hummer and co-workers [26], while the latter is more familiar and includes, for instance, Go models that use a single bead to represent each amino acid and native-centric potential functions [27, 28]. These CG models are potentially better connected to experimental observables, such as single molecule FRET data, although establishing the connection is not always straightforward given the potentially complex kinetic networks of a biomolecular system. Similarly, how to best connect atomistic and CG models and reconcile the results remain a challenging frontier [25].

In addition to molecular simulations, thermodynamic models [29] are also powerful for establishing relationships between protein stability, flexibility and function and even for making meaningful predictions. Making connections between thermodynamic models and microscopic models is an intriguing problem that likely has an important impact on our understanding of “design rules” of functional motions in biomolecules.