Task 9 - Normal Mode Analysis

Several experimental techniques, such as X-ray crystallography, NMR and spectroscopy, can provide information on the structure and dynamics of biological macromolecules, in our case proteins. However, experimental methods are often time-consuming and do not provide a complete picture of the dynamic properties of proteins. Structural bioinformatics can complement experimental methods.

Molecular dynamics (MD) simulations provide invaluable insight into protein dynamics considering the full range of harmonic and anharmonic motions at the atomic level. However, as you have found out by now, MD simulations are computational expensive. Typical simulations sample conformational motions on the nanosecond timescale.

Normal mode analysis (NMA), on the other hand, has been used successfully to determine and investigate large global motions of proteins. In NMA, the protein is modeled as a harmonic system oscillating around a stable equilibrium. Anharmonic motions are neglected. The low-frequency modes correspond to collective motions of the complete protein.
The first NMA of a protein used an all-atom representation of bovine pancreatic trypsin inhibitor (BPTI). You can have a look at the papers here:
Brooks & Karplus 1983
Go, Noguti & Nishikawa 1983

Elastic network models (among them Gaussian and anisotropic network models) greatly reduce the memory requirements for NMA. In 1996, Monique Tirion introduced this simplified model that was further developed by several others during the next years. You can find a short overview here:
http://mmb.pcb.ub.es/FlexServ/help/NMA.php
Two original papers can be viewed here:
Tirion 1996
Hinsen 1998

Introductory talks

The talk gives an introduction to normal mode analysis:

Tasks and questions

In this task you will analyze your protein structure using elastic network models. You will use two servers to calculate the normal modes:

WEBnm@ and ElNemo

For each server, calculate and analyze the lowest five (to ten) normal modes. If possible (for ElNemo), use a cutoff for C_α atom pairs of 10 Å. Note: ElNemo reads only the ATOM record from the PDB file. If your protein has a ligand which is given as HETATM, you need to change this to ATOM, if it should be accounted for in the normal mode calculation.

• What information do the different servers provide?
• How are the normal modes calculated, that is from which part of the structure? How many normal modes could in principle be calculated for your protein without any cutoff.
• Visualize the modes (provided by server or using for example PyMol or VMD) and describe what movements you observe: hinge-movement, “breathing”…
• Which regions of your protein are most flexible, most stable?
• Can you identify domains for your protein? Compare to the CATH, SCOP and Pfam domains of your protein.
• Can you observe notable differences between the normal modes calculated by the different servers?
• For WEBnm@ try the amplitude scaling and vectors option.
• When your MD simulations are finished, compare the lowest-frequency normal modes with your MD simulation using visualization software, e.g. PyMol or VMD. Can you observe different movements or similar dynamics? If possible, compare an overlay of the lowest-frequency modes to your MD simulation. You can superimpose the normal modes for example in VMD.
• What are the advantages and disadvantages of NMA compared to MD?

Here are some other servers:

Anisotropic Network Model web server

oGNM – Gaussian network model

NOMAD-Re

References

Nathalie Reuter, Konrad Hinsen & Jean-Jacques Lacapère. (2003) Transconformations of the SERCA1 Ca-ATPase: A Normal Mode Study. Biophys J 85(4): 2186–2197.

WEBnm@

ElNemo

VMD

PyMol