Normal modes of biological macromolecules: methods and applications
- Yves-Henri Sanejouand (Université de Nantes and CNRS, France)
- Konrad Hinsen (Centre de Biophysique Moléculaire (CNRS), France)
- Marc Delarue (Institut Pasteur, France)
It has been well recognized that dynamics is essential for protein function. Local motions such as amino-acid sidechain fluctuations can be studied with various experimental techniques, or predicted in silico by molecular dynamics (MD) simulations. Global motions are more challenging. X-ray cristallography can for instance be used for determining the endpoints of a motion, but the process can prove slow. In principle, MD simulations could also be used for predicting these motions. However, such motions occur on a timescale which is well over the microsecond, that is, the timescale that can be reached nowadays, using standard supercomputers. Apart from MD, another in silico technique has been used for predicting global motions, namely, normal mode analysis (NMA), which has the advantage of being much faster. Actually, NMA was among the very first in silico techniques used for trying to predict the motions of a biomolecule (R. Levy, D. Perahia & M. Karplus, 1982; T. Noguti & N. Go, 1982). At the time, the technique was limited by the amount of computer memory required, but the development of dedicated methods (L. Mouawad & D. Perahia, 1993) allowed its application to large systems and to the demonstration that the global motions provided by NMA are often similar to protein functional motions. It was then shown that comparable results can be obtained even when the protein is described with highly simplified models (F. Tama & Y.H. Sanejouand,
2001; M. Delarue & Y.H. Sanejouand, 2002), such as elastic network models (M. Tirion, 1996; I. Bahar, A.R. Atilgan & B. Erman, 1997; K. Hinsen, 1998). The workshop will noteworthy bring together the main developpers of these models (ENMs).
Some recent reviews: W. Zheng & H. Wen 2017, Curr. Op. Str. Biol. vol.42, p24; S. Mahajan & Y.H. Sanejouand 2015, Arch. Bioch. Bioph. vol.67, p59; I. Bahar & col. 2010, Ann. Rev. Biophys. vol. 39, p23.
Main challenges and topics to be discussed:
*) Studying very large systems: Modern structural techniques can yield structures of very large systems, like the ribosome. Through coarse-graining, NMA can be used for studying systems of any size. However, at the moment, there is no consensus on how to best coarse-grain a system on a large scale, without loosing its most significant characteristics.
*) Comparative dynamics of related proteins: A key aspect of modern biology is to understand structure and function in the context of evolution. For proteins, this is routine work for structure, because appropriate measures for structural similarity are well established. Measuring the similarity of the dynamics of related proteins is a more difficult task. Several methods have been proposed, but they still need to be assessed.
*) Managing flexibility in integrative structural biology: In the past, most protein structures have been determined by X-ray crystallography and NMR. Integrative structural biology solves protein structures by combining data from different experimental sources, which individually would be insufficient. For data obtained from proteins in solution (SAXS, CD, ...) or from nonhomogeneous assemblies (electron microscopy), the flexibility of the proteins must be taken into account when combining the different types of input data. The use of normal modes is a very promising approach which has not yet been exploited to its full potential.
*) Flexible docking: Most proteins function by interacting with other partners, some proteins like calmodulin interacting with hundreds of other ones. This represents a challenge for the field of structural biology, as the size of the interactome is at least one order of magnitude above the number of genes of an organism. To address this issue, many in silico docking methods have been proposed, but they are working well only when both partners remain rigid enough upon association. It has already been shown that NMA has the potential of predicting motions occuring upon association (E. Lindhal & M. Delarue, 2005; M.Sternberg and col., 2008; I.H. Moal & P.A. Bates, 2010) but significant improvments of the methods are still needed.