TMV segclassreconstruct results

[Benoît Zuber]

Hello. 

I ran segclassreconstruct on a class average obtained from the 4 TMV images of the tutorial. Initially I used the same symmetry pair grid as in the tutorial, but with images binned by a factor 4 for the sake of speed. There were correlation peaks but it was not quite the same pattern as expected from figure 7 of the tutorial. I thought this is maybe due to the binning, so I worked with a class average that had a binning factor of 2. Then for it the test to run faster, I restricted the parameter grid as follows:

Helical rise/rotation or pitch/number of units per turn choice = Pitch/unit_number

Range of helical rise or pitch search in Angstrom = (22.5, 23.5)

Range of helical rotation in degrees or number of units per turn search = (15.0, 17.0)

Increments of helical symmetry steps in Angstrom or degrees = (0.1, 0.01)

Looking at the average amplitude correlation in the grid database, which I have attached, we can see expected peaks at 15.66 and 16.34 units per turn. However there are stronger peaks at 15.49, 15.51, 16.49, and 16.51 and much stronger peaks at 15.02 and 16.98. Has anyone made this observation before? Am I doing something wrong?  I have also attached the parameter file used for segclassreconstruct. 

Thanks in advance

Ben

[Carsten Sachse]

Dear Benoit,

I think this can be explained if you have not provided a class that lies in the image plane. Does your class average look symmetric? You can also check the phase differences of the layer lines using SEGCLASSLAYER to check whether the phase differences change over resolution or they remain constant. This will tell you whether the class is in the plane or significantly out of plane.

Best wishes,

Carsten

 

Thanks in advance

Ben

[Benoît Zuber]

Dear Carsten

Thank you for your reply! Indeed I mistakenly selected a class that did not look symmetric for the grid I uploaded. In other cases, the real space image of the class I selected looked symmetric. Nonetheless I ran segclasslayer and all the classes that to me look symmetric and with features over the central cavity show a progressive increase of phase difference with resolution. I attach a screenshot of the analysis of such a class. I guess this is not what we want, right? Do you have any suggestion of how I could fish better class averages. I asked for 10 classes in most segmentations, as well as for 15 classes in the last case. Would you recommend to go for more class averages?

Best wishes
Benoit

[Benoît Zuber]

sorry I meant 10 classes in most classifications...

[Benoît Zuber]

More reading on helical theory and analysis of class averages indicate that these class averages that looked good to me are actually not centred. Probably that's what you meant by symmetric. I'll play more to get them centred.
Best
Benoît

[Carsten Sachse]

Dear Benoit,

You are correct. There are two prerequisites for SEGCLASSRECONSTRUCT. First, the class average should not exhibit any out of plane tilt.  Second, the class average needs to be perfectly centered.  If the features of the left and the right hand side of the helix differ significantly, this indicates that the helix is out of plane. In fact, this property can be used to determine the absolute hand by tilting helical specimens- for more details refer to Finch, J. T. (1972). The hand of the helix of tobacco virus. Journal of molecular biology, 66(2), 291–294. You can check the centering of your helix using SEGCLASSEXAM by the 1D helical projection or helix width profile.

I hope this make sense to you now.

Best wishes,

Carsten

[Benoît Zuber]

Dear Carsten
Thank you four reply. I'll check the reference. Of course it is trivial to shift the class average to centre it based but I was wondering if there was any automated procedure for doing that implemented in spring.
Best wishes
Benoît

[Carsten Sachse]

Dear Benoit,

Indeed, the way we tend to use SEGMENTCLASS is to run it two times. The first time we generate classes de novo. The second time, we take the clean and centered classes and provide them as a reference to rerun SEGMENTCLASS. This way, you will have them centered for SEGCLASSRECONSTRUCT.

Best wishes,

Carsten