Normal mode analysis HEXA

From Bioinformatikpedia
Revision as of 08:23, 17 August 2011 by Uskat (talk | contribs) (Results)

Webnma

General information

The WEBnma (Webtool for Normal Mode Analysis) is a webtool which allows the user to run a normal mode analysis of its protein. It is publishing available [here]. This tool offers two different analysis modes (single analysis and comparative analysis). We used the single analysis mode. Therefore, we upload our PDB file, but it is also possible to paste only the PDB id of the protein. Furthermore it is possible to declare which chain of the protein should be analysed. We pasted our e-mail adress and used the standard parameter for the analysis.
As a result we got a plot with normalized squared atomic displacments for each model. Furthermore we got pictures of the motion of the protein, which can be seen in the result section.

Results

We analysed the five models with the lowest energy values. Webnma calculates fourteen different models with following energy values:

Mode Index Deformation Energy
7 1041.88
8 1318.21
9 1738.30
10 2841.31
11 3135.09
12 4100.18
13 3911.06
14 5337.50
15 5741.69
16 6513.85
17 6081.05
18 6882.96
19 7514.14
20 7943.67

We took the first five models (7, 8, 9, 10, 11), because they have the lowest energy of all models.

Here you can see the normalized squared atomic displacments of our models:

model 7 model 8 model 9 model 10 model 11
normalized squared atomic displacments of model 7
normalized squared atomic displacments of model 8
normalized squared atomic displacments of model 9
normalized squared atomic displacments of model 10
normalized squared atomic displacments of model 11


Here you can see the motion of our analyses.

model 7 model 8 model 9 model 10 model 11
normalized squared atomic displacments of model 7
normalized squared atomic displacments of model 8
normalized squared atomic displacments of model 9
normalized squared atomic displacments of model 10
normalized squared atomic displacments of model 11

Discussion

If we look at the squared atomic displacments of our models we can see that almost every model has a peak at the same place.

Here is an overwiew, where are the peaks of the models:

model 7 model 8 model 9 model 10 model 11
position of the peak value of the peak position of the peak value of the peak position of the peak value of the peak position of the peak value of the peak position of the peak value of the peak
250 3.0 10 0.8 10 0.7 250 8 20 0.6
290 1.5 12 0.8 15 0.7 220 0.8
300 1.5 60 1.3 50 0.8 310 0.9
310 1.5 90 1.0 110 1.8 315 0.9
110 0.6 190 0.6 350-380 ~1.2
115 0.6 250 0.7 405 1.5
250 2.0 310 1.2 500 0.9
300 1.0

As listed in the table above, each model has a different amount of peaks and also the values differ extremly. Therefore, if you look at the plots you have to keep in mind, that the axis are not equal by each model. In out case, model 8 has the most peaks, whereas model 10 has only one peak, but this peak has the highest value of each peak in the models.
Model 7 has four points in the structure where the peak is very high. These points are located near by each other, which means that there is a very flexible region in the protein. If we look at the picture we can see that the highest helix is very flexible, which is probably the region of the protein where the peaks are very high. The rest of the protein moves, because of the hugh movement of this helix.
Model 8 has a lot of flexible regions, which are not located near by each other. Therefore, the protein moves on different locations. This means, that there are several independent movements and not only one movement which force the rest of the protein to move. If we have a look at the picture, we can see that there are some regions which move indenpently, as the two helices on the right botton or also the highest helix, which move from right to left and back.
Model 9 has also a lot of peaks in the plot. The peaks are very similar to them of model 8, which means these two models should move in a very similar way. This is the case, as can be seen on the pictures. Model 9 shows lower movement, than model 8, but the direction of the movememnt and the regions are the same. This can be explained, because model 8 has some more regions which are flexbile and therefore should show a stronger movement.
Model 10 has only one very high peak at the region around residue 250. By a look at the picture it can be seen that the highest helix shows an extensive movement, which force the rest of the protein to move with it. The movement is similar to them of model 7, but more concentrated to the helix at residue 250. The last model, model 11 has also a lot of peaks and one extremly flexible region. The protein moves around the ring. The complete protein wobble around this ring, which is a similar to the movement of model 8, but here this movement is more pronounced.

In general, all models show relativly similar movements. Especially, the helices at the bottom right and at the top are very flexible, which is consistent in each model.

The Normal mode analysis is just a coarse approximation of the movememnt of the protein. But we think that its pretty sure, that these three helices are really flexible, because this is the consensus of all of the different models. Therefore, we suggest, that the protein wobbles around the ring with movements of these helices.

El Nemo

General information

ELNemo is a Webinterface to compute Normal Modes by using The Elastic Network Model. The Webserver is available [here].
As input we uploaded our PDB file. Optional it is possible to assign a job name. We changed the parameters according to the task description. As output we got a fluctation matrix for each model and also three different pictures of the protein motions from different perspectives.

Results

Here you can see the CA distance fluctuations for the different modes:

This matrix displays the maximum distance fluctuations between all pairs of CA atoms and between 
the two extreme conformations that were computed for this mode (DQMIN/DQMAX). Distance increases are
plotted in blue and decreases in red for the strongest 10% of the residue pair distance changes. 
Every pixel corresponds to a single residue. Grey lines are drawn every 10 residues, yellow lines
every  100 residues (counting from the upper left corner).


model 7 model 8 model 9 model 10 model 11
Fluction matrix between the two different modes from model 7
Fluction matrix between the two different modes from model 8
Fluction matrix between the two different modes from model 9
Fluction matrix between the two different modes from model 10
Fluction matrix between the two different modes from model 11

El Nemo provides three different animated gif to the user, which show the motions of the protein. Furthermore, we created another animated gif. Therefore, in this case we have four different anmiated gifs. We listed all gifs for one model in on line.

Model 7:

picture from us picture 1 from ElNemo picture 2 from ElNemo picture 3 from ElNemo
Picture of the motions of Model 7 from El Nemo, which is generated by us
Picture 1 of the motions of Model 7 from El Nemo, which is generated by the program
Picture 2 of the motions of Model 7 from El Nemo, which is generated by the program
Picture 3 of the motions of Model 7 from El Nemo, which is generated by the program

Model 8:

picture from us picture 1 from ElNemo picture 2 from ElNemo picture 3 from ElNemo
Picture of the motions of Model 8 from El Nemo, which is generated by us
Picture 1 of the motions of Model 8 from El Nemo, which is generated by the program
Picture 2 of the motions of Model 8 from El Nemo, which is generated by the program
Picture 3 of the motions of Model 8 from El Nemo, which is generated by the program

Model 9:

picture from us picture 1 from ElNemo picture 2 from ElNemo picture 3 from ElNemo
Picture of the motions of Model 9 from El Nemo, which is generated by us
Picture 1 of the motions of Model 9 from El Nemo, which is generated by the program
Picture 2 of the motions of Model 9 from El Nemo, which is generated by the program
Picture 3 of the motions of Model 9 from El Nemo, which is generated by the program

Model 10:

picture from us picture 1 from ElNemo picture 2 from ElNemo picture 3 from ElNemo
Picture of the motions of Model 10 from El Nemo, which is generated by us
Picture 1 of the motions of Model 10 from El Nemo, which is generated by the program
Picture 2 of the motions of Model 10 from El Nemo, which is generated by the program
Picture 3 of the motions of Model 10 from El Nemo, which is generated by the program

Model 11:

picture from us picture 1 from ElNemo picture 2 from ElNemo picture 3 from ElNemo
Picture of the motions of Model 11 from El Nemo, which is generated by us
Picture 1 of the motions of Model 11 from El Nemo, which is generated by the program
Picture 2 of the motions of Model 11 from El Nemo, which is generated by the program
Picture 3 of the motions of Model 11 from El Nemo, which is generated by the program

Discussion

This method provides a fluctation matrix and also some pictures of the motion. The fluctation matrix visualize if two residues come closer together (red) or are more distant (blue) during the motion. So it is possible, to see which residue of the protein behaves in which way. Furthermore, we got some pictures of the motion in different perspectives, which it made easy to get an imagination of how the protein moves.

In the first model (model 7) there almost the same amount of residues which come closer together and are more distint. If we have a look at the picture, we can see, that the whole protein moves. In the first perspective of ELNemo we can see, that the protein pulsate. Especially one helix shows an up and down movement, which cause the rest of the protein motion. Only the ring in the middle of the protein is relativly stable and shows only low motion.
The second model (model 8) has a different fluctation matrix. At the beginning of the protein the resides are more distant, whereas, at the end of the protein the residues come closer together. In the pictures we can see that, as before, the upper helix shows a strong motion, but also the lower sheets show stronger motion than in the model before. In general this model shows more motion than before, which can also be seen at the ring, which moves clearly more than in model 7. Again, as before, the protein seems to pulsate.
Although the fluctation matrix of model 9 is different of the matrix of model 8, the motion seems to be very equal. In the matrix it can be seen, that the residues just com closer together. There are almost no residues, which get a greater distance to the residues than in their QMIN state. By a look at the pictures it can be seen that the motion of model 9 and model 8 is quite similar. Model 9 only shows a little less motion in the first helix, but the general movement ist the same. So therefore again, the protein seems to pulsate around the ring.
In model 10 almost all residues come closer together or keep their distance. This matrix is really different from the matrices before and therefore, we can suggest, that the protein motion is different of the already discussed models. By a look at the picture we can see that this assumption is correct. The protein shows a more complex motion than the models before. The protein still moves the helix at the top, but the motion is very small. Otherwise, the complete protein puslate, but the pulsation is more distinct than before. The whole protein seems to pulste twice, which could not be seen in the other models.
The last model has a fluctation matrix which is most similar to them of model 10. There are less residues which come closer together, but as before, there are no residues which enlarge the distance. So we suggest, that the motion has to be similar to them of model 10, although the motion should not that pronounced than by model 10. If we look at the picture, we can see, that this is true. The protein shows less motion and especially the pulsation seems to be low. Only if we look at picture two, we can see, that the protein has a hugh pulsation, but the direction is different to the models before.

In general, we can say, that the ring in the middle of the protein is stable and the protein pulsate around them. The different models show different possible occurences of this motion. The top helix is a very flexible one, which has a motion in all different models. Futhermore, each model shows pulsation, so we can suggest, that the upper helix moves and the protein pulsate around their center, which is a ring in our case. All models show similar motions, especially the motion of the upper helix and also of the lower sheets, which are also flexible in each model. So therefore, we can get a coarse imagination of how this protein moves and which parts of the protein are the most important ones for motion.

Anisotropic Network Model web server

General information

The Anisotropic Network Model web server can be found [here].
In the web interface, we uploaded our PDB file and chose the chain for which the analysis should be done. Furthermore, we changed the parameters according to the task description.
As an output we a plot of the real occuring and calculated B-factor values. Furthermore, we got a picture for each model with the protein motion. Very nice in this case is, that there are arrows in the picture which shows the direction of the motion. So therefore, it is very easy to reconstruct the motion directions.

Results

Here you can see the B-factor distribution of the real occuring B-factors (black) and the calulated B-factors (blue).

model 1 model 2 model 3 model 4 model 5
Distribution of the B-factors and the calculated B-factors in Model 1
Distribution of the B-factors and the calculated B-factors in Model 2
Distribution of the B-factors and the calculated B-factors in Model 3
Distribution of the B-factors and the calculated B-factors in Model 4
Distribution of the B-factors and the calculated B-factors in Model 5

In the next table, the motion of the protein of the different models is shown:

model 1 model 2 model 3 model 4 model 5
Calculated motion of the ANM model 1
Calculated motion of the ANM model 2
Calculated motion of the ANM model 3
Calculated motion of the ANM model 4
Calculated motion of the ANM model 5

Discussion

This method provides a distribution of the real occuring B-factors and of the calculated B-factors. By comparing both values we can see which parts of the protein are flexible. A high B-factor value means a high flexbility of this residue. If we look at the plots in detail, on the first view all values seem to be very similar. Each model has four very high peaks at the same place, but at the rest of the protein the values differ between the different models. All models have also big peaks at the beginning of the protein, which was expected, because the beginning and the end of a protein are often very flexbile (therefore, most methods which predict disordered regions predict them at the beginning or the end of the protein).

The first model has only one loop which shows strong motion and therefore the rest of the protein follows the motion of this loop. It seems that the whole protein contract and strecht. This motion is caused by the motion of the flexible loop.
The second model shows a complexer motion, with not only one motion center. There are two contrary motions. As before, the loop moves from right to left. But at the other side of the protein, there is a second loop which also moves forward and backward. These two loops move in the same time to the same direction. Therefore, it seems that the protein build a deeper curve by moving the two loops. The rest of the protein follows the motion of these two loops. Only the ring does not change its structure in a very obvious way.
In the third model there are again two loops which moves. They are the same loops as before. But this time, the loops move contrary. If one loop moves forward, the other loop moves backward. Therefore, the rest of the protein has not to move that much as in model 2. The ring again is very stable and does not change its structure.
The forth model has only one loop which moves. But in this case not only the loop moves, but also the associated helix. This cause a stronger movement of the whole protein as in model 1. This is the first case in which the ring also moves itselfs. So in this case the loops, which build the ring, come closer together.
The last model, we look at, has three different motions. The two loops from model 2 and 3 moves again. But there is also a helix which moves. Compared to the other models, this model shows the strongest motion and the whole protein seems to be very unstable. Also the ring shows a lot of motion.

In general there are two parts of the protein which seems to be in almost all models the center of the motion. The last model is very unstable and therefore, we do not suggest that this is the real motion of the protein. We suggest, that especially the two loops are the center of the motion and the rest of the proteins follows the motion. It was not possible to decide if both loops moves in the same or in contrary directions, because these models are just coarse approximations of the protein motion and both cases could be the real motion. But in sum we can suggest, that the protein has two loops which act as motion center, the ring does not move that much and keeps its structure and the rest of the protein follows the motion of the two motion centers.

oGNM – Gaussian network model

General information

The Gaussian network model (oGNM) web server can be found [here].

The oGNM is an online server which can calculates the essential dynamics of PDB structures and of user-modified or unrealesed structures. The method is based on an Elastic Network (EN) model called the Gaussian Network Model (GNM). The online server was an enlargment of the already existing database iGNM which also uses the GNM and which contains already a lot of PDB structures. The advantages of the server compared to iGNM or other normal mode analysis tools is that the calculation is very fast and a result is received in a few seconds. Furthermore, it is not limitied in size which means it is not necessary to use really small structures like in other normal mode analysis methods.
For the calculation the method uses a network model that represents the biomolecular structure. This network can be identified by the alpha-carbon atoms and some other selected atoms on nucleotides. Furthermore, they assume that the fluctuations of the nodes are isotropic and Gausian. Therefore, there are different parameters which can be selected before the calculation: number of model nodes to represent a nucleotide (1 or 3) and the interaction cutoff distance for nucleotide and amino acid pairs (the c-alpha cutoff is the average of this two cutoffs).
The method delivers plenty of different output files, like the Eigenvalues and the Kirchoff matrix which play a role in the calculation as well as the 20 slowest modes which is part of the results. Furthermore, there are three main representive results which displays the fluctuation and the mobility of the structure as well as cross-correlation plot between the desired modes.

(Sources: [Paper] and [Webserver])

Results

Full diversity of the received output can be seen [here]. For the analysis of the result we decided to display only the three main visualizations.

Here you can see the mobility profiles. This plots can be used for the comparitive analysis of the different mode profiles. On the x-axis are the residues of chain A and the y-axis displays the according fluctuation. The first five pictures display the modes 7-11 on their own and the last picture displays all modes together.

mode 7 mode 8 mode 9 mode 10 mode 11 all modes
Fluctuation of mode 7
Fluctuation of mode 8
Fluctuation of mode 9
Fluctuation of mode 10
Fluctuation of mode 11
Fluctuation of the modes 7-11

In the next picture display the mobility of the different modes. The mobility goes from blue to red, where blue means little mobility and red means high mobility.

mode 7 mode 8 mode 9 mode 10 mode 11
Calculated mobility of mode 7
Calculated mobility of mode 8
Calculated mobility of mode 9
Calculated mobility of mode 10
Calculated mobility of mode 11

Here you can see the cross-correlation plot of mode 7-11.

Cross-correlation plot of mode 7-11

Discussion

NOMAD-Ref

General information

Results

model 1 model 2 model 3 model 4 model 5
Histogram of the cRMS for model 1
Histogram of the cRMS for model 2
Histogram of the cRMS for model 3
Histogram of the cRMS for model 4
Histogram of the cRMS for model 5


model 1 model 2 model 3 model 4 model 5
model 1
model 2
model 3
model 4
model 5

All-atom NMA using Gromacs on the NOMAD-Ref server

General information

Results

600K

Here you can see the motion of 1BPT at temprature 600K.

model 7 model 8 model 9
normalized squared atomic displacments of model 7
normalized squared atomic displacments of model 8
normalized squared atomic displacments of model 9

2000K

Here you can see the motion of 1BPT at temprature 2000K.

model 7 model 8 model 9
normalized squared atomic displacments of model 7
normalized squared atomic displacments of model 8
normalized squared atomic displacments of model 9

Comparison to an Elastic Network Calculation

Next, we want to compare the result of All-atom NMA to an elastic network calculation (NOMAD-Ref).

model 7 model 8 model 9
normalized squared atomic displacments of model 7
normalized squared atomic displacments of model 8
normalized squared atomic displacments of model 9