Difference between revisions of "Fabry:Normal mode analysis"
(→Comparison and Conclusion)
|Line 290:||Line 290:|
== Comparison and Conclusion ==
== Comparison and Conclusion ==
=== Comparison WEBnm@ and elNémo ===
Over all the modes calculated by both methods look similar (see [[Fabry:Normal_mode_analysis#Mode_Visualization|Mode Visualization WEBnm@]] and [[Fabry:Normal_mode_analysis#Mode_Visualization_2|Mode Visualization elNémo]]), especially mode 7 through 9. There are two main differences between the results of WEBnm@ and elNémo. First, in the elNémo algorithm it is possible to consider included hetero atoms in the analysis, which apparently has an impact on the normal modes, because most modes of 3HG3 have an opposite direction when compared in the two methods. The second difference is the performed analyses and the given output. In our opinion although elNémo's results are easier to interpret, due to the better usability and the posibility to download the results of all steps, we would prefer to use WEBnm@ or a combination of both servers.
=== Domains ===
=== Domains ===
==== CATH ====
==== CATH ====
Latest revision as of 19:33, 16 August 2012
Fabry Disease » Normal mode analysis
For further information on the execution, please refer to our Journal
- 1 Introduction
- 2 WEBnm@
- 3 ElNémo
- 4 Comparison and Conclusion
- 5 References
Maybe one of the first questions that can be asked in this task is, why we use low-frequency normal modes. This is explained in the paper of Marc Delarue and Philippe Dumas<ref>Marc Delarue and Philippe Dumas On the use of low-frequency normal modes to enforce collective movements in refining macromolecular structural models, Proc. Natl. Acad. Sci. (USA), 101, 6957-6962 (2004)</ref>, where they claim, that "many of the structural transitions (...) can be explained by just a few of the lowest-frequency normal modes". The normal modes can be used to generate the general motion of a system by superposition them. Thus we could in principle infer from our analysis in this task how the alpha-galactosidase A, which we examine hydrolyses the terminal alpha-galactosyl moiety of its substrate<ref>Normal mode http://en.wikipedia.org/wiki/Normal_mode, July 5th, 2012</ref>.
We decided to use the structures 3HG2 and 3HG3, which represent the human α-Galactosidase catalytic mechanism with empty active site and substrate bound, respectively (see <xr id="fig:bindSite"/>). From this, we hope to getter an insight into the mechanism and a possibility to compare normal modes and the behaviour of the molecule.
Possible normal modes:
The structure 3HG2 has 781 C-alpha atoms in its pdb file, thus 781*3 - 6 = 2337 normal modes could be calculated in principle for this structure without any cutoff by elNémo. Since the structure has a total of 6765 atoms, 20289 NMs could be calculated by WEBnm@.
The structure 3HG3 has 793 C-alpha atoms in its pdb file, thus 793*3 - 6 = 2373 normal modes could be calculated in principle for this structure without any cutoff by elNémo. Since the structure has a total of 7537 atoms, 22605 NMs could be calculated by WEBnm@.
WEBnm@ <ref>Hollup SM, Sælensminde G, Reuter N. WEBnm@: a web application for normal mode analysis of proteins BMC Bioinformatics. 2005 Mar 11;6(1):52 </ref> claim to administer simple and automated computation of low-frequency normal modes for proteins as well as their analysis in order to clarify if it is beneficial to perform a complete study on the protein in question.
The server calculates Normal Modes with the help of the MMTK package <ref>Hinsen K, The Molecular Modelling Toolkit: a new approach to molecular simulations, J Comput Chem, 21:79-85, 2000</ref>, which is an Open Source program library for molecular simulation applications. A C-alpha force field <ref>Hinsen K, Petrescu AJ, Dellerue S, Bellissent-Funel MC, Kneller GR, Harmonicity in slow protein dynamics, Chemical Physics, 261:25-37, 2000</ref> is used and only these C-alpha atoms are used, but with a weight assigned that corresponds to the masses of the whole residue they represent.
The server provides a bunch of analysis tools and all results can be downloaded without any problems. The tools are:
- deformation energies of each mode
- calculation of normalized squared atomic displacements
- calculation of normalized squared fluctuations
- interactive visualization of the modes using vector field representation or vibrations
- correlation matrix
Deformation energies and eigenvalues
In <xr id="fig:eigen3HG2"/> and <xr id="fig:eigen3HG3"/> the Eigenvalues for the first 50 modes for both examined structures are plotted. The increase of the values shows a decrease of amplitude of motion in the modes, since there is an invers relationship between the Eigenvalues and the amplitude. Hence, mode 7 has the highest amplitude.
Since the eigenvalues correspond to the frequencies and a low frequency tends to describe a global movement of the protein our assumption is confirmed, that the lower modes express global movement, while higher modes rather show many smaller local movements. <ref>Normal Mode (Harmonic) Analysis http://cmm.cit.nih.gov/intro_simulation/node26.html, August 14th, 2012</ref>
In figure <xr id="fig:avEn"/> the average deformation energies of the lowest 14 modes of both catalytic mechanism are compared. For most modes the energies of the modes for 3HG2 are bigger than those for 3HG3. This means, that the amplitude of the motion is in general slighty lower in 3HG2 than in 3HG3 (see also Deformation Energies Table for the values of the average energies).
In this section, we want to visually inspect the motion of the protein in the 6 smallest modes that were identified by WEBnm@.
For a description of the modes of the molecules 3HG2 and 3HG3 see <xr id="tab:webnma_3hg2" /> and <xr id="tab:webnma_3hg3"/>, respectively. Modes 7 through 9 are similar for both molecules and mode 11 of 3HG2 corresponds to mode 10 in 3HG3. Although mode 7 of 3HG2 moves outwards, while mode 7 of 3HG3 moves inwards, making them moving in opposite directions. Of course, all similar modes can be explained in both states of the α-galactosidase catalytic mechanism, especially, mode 11/10, where one active site is in the process of releasing the substrate and one is binding it. Left is to compare both modes 12, where 3HG2's looks like closing in on the substrate that is to be bound, while 3HG3's mode 12 could be the beginning of releasing the hydrolized sugar.
For the modes 3HG2 10 and 3HG3 11 we do not have an explanation, why they could explain the function of our protein.
Concluding, it can be said, that the protein α-Galactosidase is rather rigid in most parts, except for the residues that connect chain A and B and the big and small part of each chain.
Atomic Displacement Analysis
In <xr id="fig:atomDisplacement3HG2"/> and <xr id="fig:atomDisplacement3HG3"/> the square of the atomic displacements of the C alphas of the examined structures are shown. These are normalized in a way, such that the sum over all residues is equal to 100. With these plots we can find out, which regions are displaced most, i.e. which move the most; this is shown by the highest values. It is recommended to look for clustered peaks, which identify significantly big regions. Local flexibility (a single peak) is of less importance.
In both figures, chain A and B are colored different. From this we can see, that although in <xr id="tab:webnma_3hg2"/> and <xr id="tab:webnma_3hg3"/> the motion of both chains in a mode looks alike, but in general it is not perfectly equal or even differs a lot. A good example for a significant variation is mode 7 of the structure 3HG2 (see <xr id="fig:atomDisplacement3HG2"/>, upper left). While the first part of both chains (approximately until position 200) behaves similiar, except for a different amplitude, the second half differs with chain B showing much more movement. This action can be oberserved in <xr id="tab:webnma_3hg2"/> only after a very close inspection.
For the dimer it seems to be easier to act different when no substrate is bound, since the atomic displacements of the chains in the modes of 3HG3 seem to be much more alike than those of 3HG2.
All in all, the substrate binding site itself (residue 203 to 207) seems to be rather ridgid, except for maybe mode 10 in 3HG3, while the ends of both chains (the last 50 residues) are fairly flexible. This can best be seen in both modes 12.
Fluctuation is the sum of the atomic displacements of each C alpha atom in each non-trivial mode weighted by the inverse of their corresponding eigenvalues. These are normalized in a way, such that the sum over all residues is equal to 100. The fluctuations of the structures 3HG2 and 3HG3 are shown in <xr id="fig:fluctuation_3HG2"/> and <xr id="fig:fluctuation_3HG3"/>, respectively.
The plots support our previous assumption, that the chains of the substrate bound structure 3HG3 act much more similiar than those of the structure with an empty active site, since the overlap is almost perfect in the left plot of <xr id="fig:fluctuation_3HG3"/>.
Again we can observe that the binding site is an rigid island among two moderate flexible regions, which probably are responsible for opening and closing the binding pocket and the needed movement for the breake down of the bound sugar.
Towards the end of the chain more motion can be observed, which is needed for the independant movement of both chains.
In the plots in <xr id="tab:corrMatr"/> both correlation matrices of 3HG2 and 3HG3 are shown. Over all, the plots look very similar, showing a positive correlation of the first part of both chain A and B to each other and also of both second halves to each other. And a negative correlation of the first half of chain A to the secon half of chain B and vice versa. The chains among themselves are rather strong positively correlated along the less strict diagonal and in the second half of the chain and negatively correlated in the rest. This underlines our statement (see section Mode Visualization) that the protein is quite rigid, with only the connecting parts being flexible and also that as well both chains move away from each other or towards each other, as the two halves of each chain itself can move independently.
The only difference between the plot of 3HG2 and 3HG3 is the strength of the correlations, where the colors in the second plot are darker, indicating a stronger correlation and therefore a higher amplitude which we have already seen in section Average Energies.
General error. A trouble ticket has already been send on 04.07.2012 11:22
ElNémo <ref>K. Suhre & Y.H. Sanejouand, ElNemo: a normal mode web-server for protein movement analysis and the generation of templates for molecular replacement. Nucleic Acids Research, 32, W610-W614, 2004. </ref> <ref>K. Suhre & Y.H. Sanejouand, On the potential of normal mode analysis for solving difficult molecular replacement problems. Acta Cryst. D vol.60, p796-799, 2004 © International Union of Crystallography. </ref> is the web-interface to the "Elastic Network Model", which is a tool to compute the low frequency normal modes of a protein. It performs several analyses, including
- an overall normal mode analysis taking into account frequency, collectivity of atom movement, overlap of each mode with the observed conformational change (if two conformations provided) and the corresponding amplitude
- an individual normal mode analysis with animations and RMSD for each mode
- an Analysis of distance fluctuations between all CA atoms (cross plot)
- a B-factor analysis
- a comparison of a normal mode perturbed structure and a second conformations in terms of RMSD and number of residues that are closer than 3 Angstrom
Additionally you can compute combined models from two modes. Also the server claims that there is a way to download all the results at once, which only produced errors in our case.
We decided to examine the 6 lowest modes that ElNémo provides to obtain a result that is comparable to the 6 modes by WEBnm@.
3HG2: Correlation= 0.651 for 781 C-alpha atoms.
3HG3: Correlation= 0.779 for 793 C-alpha atoms.
Both correlations are according to the developers in a normal range.w
NOTE: We wanted to extract the per-residue B-factors of both structures after rerunning the NMA, but there seems to be another error at the elNémo server that now does not accept any of our (previously accepted) pdb-files.
In this section, we want to visually inspect the motion of the protein in the 6 lowest frequency modes were identified by ElNémo. For a description of the modes of the molecules 3HG2 and 3HG3 see <xr id="tab:elnemo_3hg2"/> and <xr id="tab:elnemo_3hg3"/>, respectively.
In comparison, we see a large overlap of the modes of 3HG2 and 3HG3. Mode 7 of 3HG2 combines modes 7 and 8 of 3HG3 in it, while mode 8(3HG2) is the opposite movement to mode 8(3HG3). The same applies for both modes 10, where one moves to the right, the other one to the left. Both modes 9 are the exact same. Mode 11(3HG2) is similar to mode 12(3HG3), except for the additional independant site in the latter one. The left mode of 3HG3 (mode 11) is similar to the modes 10, but again there is one part that moves independantly from the others. The only mode that cannot be observed in both conformations is mode 12(3HG2), but it might be that it is similar to a higher frequency mode in 3HG3.
The distance variation plots in <xr id="fig:FABRY_caStrain3HG2"/> and <xr id="fig:FABRY_caStrain3HG3"/> show the distance variation between successive pairs of CA atoms in the two extreme conformations that were computed for each mode (DQMIN/DQMAX). This can be interpreted as the internal movement or flexibility. In the modes 7 and 8 of both structures there is not much variation, like we expected. In the modes 9, the variation is very similar, except that chain A in 3HG2 corresponds to chain B in 3HG3 and vice versa. From the vizual inspection before we have not expected this. Therefore we think we could have made a mistake in identifying chain A and B in the pictures in the earlier section, and see also section Distance fluctuations, where the results from here are again underlined. The same applies for modes 10 and in 11 of 3HG2 and accordingly 10 and 12 of 3HG3.
Besides from that there is moderate difference in the amplitude and, as expected, fairly long stretches without significant peaks indicating rigid regions interupted by some flexible residues. Usually, a change of behaviour can be observed after about two thirds of each chains' sequence (around position 260). This confirms our finding from section Distance variation, where we saw the flexible link between the chains (around position 257 to 267) and a smaller part of the chain (position 268 to end) behaving different than the rest.
The plots in <xr id="tab:elnemo_fluct_3hg2"/> and <xr id="tab:elnemo_fluct_3hg2"/> display the maximum distance fluctuations between all pairs of Cα in both extreme conformations for this mode (DQMIN/DQMAX). If the distance decreases, this is shown in red, an increase is shown in blue. On first sight, we see that there is hardly any distance fluctuation among the chains themselves, but between chain A and B. Only in modes 12 (3HG2) and 10 through 12(3HG3) significant amount of interaction among one chain can be observed.
The plots underline our previous findings, like the overall similarity in movements and the rigidity of most parts of the molecule. Also the division of the chains into two domains can again be observed, especially in both modes 9. What becomes very obvious here is the opposite direction of the same movement in 3HG2 and 3HG3, for example in mode 8, where one is an opening and one a closing movement.
The distance fluctuations observed in mode 11 (3HG2) can be found in the plot of mode 12 (3HG3), which additionally contains one part of chain A that increases its distance to itself.
Comparison and Conclusion
Comparison WEBnm@ and elNémo
Over all the modes calculated by both methods look similar (see Mode Visualization WEBnm@ and Mode Visualization elNémo), especially mode 7 through 9. There are two main differences between the results of WEBnm@ and elNémo. First, in the elNémo algorithm it is possible to consider included hetero atoms in the analysis, which apparently has an impact on the normal modes, because most modes of 3HG3 have an opposite direction when compared in the two methods. The second difference is the performed analyses and the given output. In our opinion although elNémo's results are easier to interpret, due to the better usability and the posibility to download the results of all steps, we would prefer to use WEBnm@ or a combination of both servers.
The server divides each of the chains into two domains, Aldolase class I from position 32 to 324 and from 325 to 421 a Golgi alpha-mannosidase II, which is a "mainly beta" domain thus containing only loops and beta-sheets in our protein.
In our analyses we could observe these two domains and refered to them as "bigger and smaller part" of the chain. We were not absolutely right about the border between the two domains, thinking based on the behaviour that the break was around position 257.
No result could be obtained from this ressource.
No additional information could be found on Pfam.