[3DEM] Handedness determination by Random Conical Tilt

[Hong-Wei Wang]

**** Messages to this list are automatically archived ***
**** Please limit quoting of previous postings to the bare minimum ****

Dear all,

This is a follow-up of the problem that I brought last week. I got quite a 
few responses regarding the handedness determination of 3D reconstruction by 
TEM. Some of them are very helpful, while some show that my problem was not 
a single case. It is worth making the issue clearer.

There are several ways to determine the handedness of a structure. As I can 
think of, tomography is one method. Metal shadowing is another. They should 
work well with structures with large size or certain feature (such as 
helical particle by metal shadowing).
For most single particles, the best way is taking tilt-paired micrographs of 
the specimen and comparing the images with the 3D model. There have been 
literatures discussing this method by tilting the sample at low angles (5-20 
degrees). Two of them are:
1. Belnap et al. (1997) J. Struct. Biol. 120, 44-51
 2. Rosenthal and Henderson (2003) J. Mol. Biol. 333, 721-745

Random Conical Tilt actually incoporates this in its processing pipeline. 
The only difference is that RCT needs high tilting angles so that both 
handedness and the 3D reconstruction are obtained simultaneously. Two 
essential factors that need a lot of cautions are the micrograph orientation 
and the direction of tilt-axis in the micrographs. Failure of determining 
either one of them may cause a mirror in the reconstruction. These two 
factors are linked. If the micrograph is viewed from bottom (such as films 
with markers readable), the higher area of the specimen in the column is 
farther from the viewer; if the micrograph is viewed from top (such as CCD 
images), the higher area of the specimen in the column is closer to the 
viewer. Some people mentioned the flip of the specimen when loading the grid 
onto the holder. This has no effect on the handedness determination by this 
method.

More specifically, for the case in my reconstruction. After I picked the 
tilted pairs of particles in WEB, WEB calculted the gamma, phi, theta. They 
are for the direction of tilting axis and the tilting angles. Since WEB 
calculates these based on the geometrical coordinates of the paired 
particles in two micrographs, it cannot tell the correct sign for any of the 
three angles. Furthermore, the signs of the output values depend on the 
initial values assigned. The correct signs have to be figured out by the 
users. The rotation convention in SPIDER and WEB have to be taken into 
account as well. This can be figured out by doing some testing in WEB.
Since the marks on my micrographs are readable, I am at a position facing 
the sample from the bottom of the microscope column. In the tilted 
micrograph, area with higher defocus represents the same area of the 
specimen at lower position in the scope, thus closer to me. By this means, I 
can determine the direction of the tilting axis and the sign of tilting 
angle. In the specific example I mentioned in my last email, I made the 
mistake in this step. The GAMMA and PHI should be of a value of about 90 
degrees rather than -90 degrees. Or the THETA should be -55 degrees rather 
than 55 degrees. This led to the wrong handedness in my reconstruction.

As several people suggested, the best way to have a unambiguous handedness 
in a RCT reconstruction is to use a known sample (such as ribosome) as 
"calibrator": take tilted pair images of such sample in a microscope 
operating at certain condition (HT, magnification, tilting direction, CCD or 
film as data collector), scan the films on a scanner under certain 
convention (orientation of the micrograph, the way that the micrograph put 
in the scanner), convert the digitized images to the working format in 
certain way, calculate the direction of tilt axis with the same initial 
values (this may be just the case in WEB. Other softwares such as CTFTILT 
can assign the signs of angles correctly). All the other unknown samples 
should be reconstructed in exactly the same convention. Thus the handedness 
is determined.

I hope that I have described clear enough and it is useful to people that 
are using random conical tilt method. A lot of thanks to those who responded 
and gave suggestions to my previous questions.

Hongwei

----- Original Message ----- 
From: "Hong-Wei Wang" <HWWang at lbl.gov>
To: <3dem at ncmir.ucsd.edu>
Sent: Friday, March 16, 2007 3:57 PM
Subject: [3DEM] Handedness determination by Random Conical Tilt


> **** Messages to this list are automatically archived ***
> **** Please limit quoting of previous postings to the bare minimum ****
>
> Dear colleagues,
>
> I have experienced a dilemma in determining the handedness of a single 
> particle by random conical tilt method. I realized that there must be 
> something wrong during my reconstruction process because the final 
> reconstruction volume shows a wrong handedness compared with the available 
> atomic model strcuture. I am going through the whole process that I did. 
> Please tell me in which step I might have done wrong to mess up the 
> handedness determination.
>
> 1. I took tilt pair images of the sample in a Tecnai-12 microscope. The 
> tilt angles were 0 and +55 degrees of the goniometer.
> 2. The micrographs were recorded on films. I digitized the films in Nikon 
> Super Coolscan 8000 and converted them to spider format.
> 3. In WEB interface, I chose the "tilt-particle" command to open the 
> untilt and tilt micrograph and picked the paired particles. Here I can see 
> that the two micrographs were both in a portrait orientation with the 
> micrograph labels at the bottom. The micrograph lables are in a readable 
> orientation. The tilt axis is about perpendicular to y axis in WEB. I can 
> tell that the highly defocused area in the tilted micrograph is close to 
> the bottom. After picking the tilted pairs of particles, WEB calculated 
> out the three angles of GAMMA, PHI, THETA as about -90, -90, 55 degrees 
> respectively.
> 4. I did the MSA and MRA of the untilted particles and calculate the 
> in-plane rotation of each particle in SPIDER as ALPHA. For each class of 
> untilted particles, I generated the euler angle document file for back 
> projection reconstrudtion with three columes:
> colume one is psi = PHI (output of WEB tilt-pair pick results);
> colume two is theta = THETA (output of WEB tilt-pair pick results);
> colume three is phi  = -ALPHA - GAMMA (ourput of WEB tilt-pair pick 
> results).
> 5. I did the back-projection reconstruction of correspondent tilted 
> particles using BP RP in SPIDER using the angular document file.
>
> My guess is that the micrographs I used for processing are flipped to the 
> real projection of the single particles. I come up with an explaination as 
> follows: If I can read the lable on the micrograph in its right order, it 
> means that I am lying beneath the electron microscope facing up to look at 
> the specimen. Thus what I see is exactly the mirrored image of the 
> projection of the sample. In such a case, my reconstruction volume is a 
> mirror of the real structure.
>
> Please tell me if I am right or wrong. Any comments are very welcome!
>
> Hongwei
>
> -----------------------------------
> Hongwei Wang, Ph D
> MS 20A-355, LSD, LBNL
> 1 Cyclotron Rd, Berkeley,
> CA 94720
>
> 510-642-2222 (O)
> 510-642-8806 (FAX)
> hwwang at lbl.gov --
> For information on how to subscribe/unsubscribe/etc., send mail to
> 3dem-request at 3dem.ucsd.edu with the text "help" in the body of the
> message.
> 
--
For information on how to subscribe/unsubscribe/etc., send mail to
3dem-request at 3dem.ucsd.edu with the text "help" in the body of the
message.