Computational methods for efficient sampling of protein landscapes and disclosing allosteric regions

Methods developed toward computational exploration of protein landscapes have become standardized tools to assess biophysical experimental findings. They are also used on their own right to discover the workings of the protein as a molecular machine, potential sites of interest for protein functioning, allosteric regions in proteins, and communication pathways between different sites on a protein. With the development of reliable force fields that describe interactions in biomolecules, molecular dynamics (MD) simulations have become the prime tool for this purpose. While it is now straightforward to carry out MD simulations up to microseconds with current computers readily available to researchers, many processes of biological interest occur on several of orders of magnitudes slower timescales. Thus, the latter problems are attackable through MD by a handful of researchers that have access to the most powerful computers. Alternatively, physics-based methods to interrogate the protein energy landscape are in continuous development to circumvent this problem. In addition to opening the routes for advancement to a large number of researchers that have access to modest computational resources, they have the advantage of providing an understanding of the mechanisms that govern protein dynamics. Here we discuss network-based approaches geared toward understanding protein dynamics. These include (i) construction of residue networks which view proteins as networks of nodes connected through local interactions and (ii) construction of proteins as elastic networks whose modes of motion may be manipulated to achieve allowed conformational changes. Limitations of the methods as well as opportunities for future exploitation are described