[3dem] Membrane protein structure: orientation issue

[Christopher Aylett]

Hi Philip,

This is why single particle analysis is a local minimisation dependent on the initial model used. The assigned orientations are based on this initial input, and providing Fourier space is well-sampled and the input is within the correct local minimum, the angular assignment and calculated 3D structure will converge.

Best wishes,

Chris

****

> On 26 Mar 2020, at 07:44, Philip Köck <koeck at kth.se> wrote:
> 
> Hi again.
> 
> I believe you are right if you can be sure that you've found a single perfect side view and the symmetry is at least C3.
> In that case I would say you can just symmetrize, as you describe, and get a rough 3D model without using either top or tilted views. Note: You are using a single side view!
> 
> I don't see how a spread of different side views would help you though.
> You run into the problem that Steve pointed out.
> You don't know how the different side views are oriented with respect to each other​ unless you have at least one
> top or tilted view to lock the relative orientations.
> 
> All the best,
> 
> Philip
> 
> 
> Från: Dmitry Lyumkis <dlyumkis at salk.edu>
> Skickat: den 25 mars 2020 17:41
> Till: Philip Köck
> Kopia: Jacopo Marino; 3dem at ncmir.ucsd.edu
> Ämne: Re: [3dem] Membrane protein structure: orientation issue
> 
> If it’s a twofold rotationally symmetric object, adhered to the air-water interface along a single “side" view, then a projection along, e.g., phi=0/theta=90 samples the same Fourier plane as the second projection along phi=180/theta=90. The reconstruction  will behave in a manner that is identical to one that is composed of exclusively “top” views of a rotationally symmetric object (or, more simply, an asymmetric particle with one preferential orientation, normalizing for the number of asymmetric units). Basically,  you’ll end up hyper-sampling around one plane in both cases, except that the planes will lie along distinct axes of the transform. Both are bad cases and, in the absence of other views, will lead to a bad reconstruction. To be clear: this is assuming that  there’s only one preferential orientation along the side view of a two-fold rotationally symmetric particle, and not more. If you have 3-fold rotational symmetry, and the sample is adhered to its “side” view, projections separated by phi=120° spacing will  lead to sampling of Fourier planes separated by phi=60° (due to symmetry). Effectively, you add two additional planes. Assuming you have enough particles, and there’s a bit of spread in the phi angle, in most practical cases, the reconstruction of a 3-fold  rotationally symmetric object will be complete (or nearly so). The higher the rotational symmetry, obviously the better. 
> 
> 
> To answer the original question, adding top views for your membrane protein will be negligible in the reconstruction. Looks like you should already get a very nice map from the current data. Give it a try. 
> 
> 
> Best, 
> 
> 
> Dmitry 
> 
> 
> 
> 
> 
> 
> 
> 
> Dmitry Lyumkis
> Assistant Professor
> The Salk Institute for Biological Studies
> 10010 North Torrey Pines Road, La Jolla, CA 92037
> T: 858-453-4100 x1155
> E: dlyumkis at salk.edu
> https://lyumkis.salk.edu
> 
> 
> 
> 
> 
> 
> 
> 
> 
> On Mar 25, 2020, at 12:06 AM, Philip Köck <koeck at kth.se> wrote:
> 
> 
> 
> 
> _______________________________________________
> 3dem mailing list
> 3dem at ncmir.ucsd.edu
> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem