Skip to Content

How to flip orientations in frealign?

11 replies [Last post]
xzhang2005
Offline
Joined: 21 Nov 2012

Hi,

I collected two datasets of the same sample, and I first did reconstructions from the dataset #1. But when I used the final reconstruction from the dataset #1 as the model for the dataset #2, I found that the reconstruction from the #2 dataset has a flipped (wrong) handness! Is there any way to flip back the handness during refinement with frealign? Set "FLIP=T" in mparameters? Thanks much.

Xing

xzhang2005
Offline
Joined: 21 Nov 2012
In my case, the model from

In my case, the model from dataset # is around 3.1a resolution, and I filtered it back to 4a as the referecne model to align/refine the datset #2 binned 2x (nyquist limt=4a). So, I think there should be clear hand info in the 3D model and I did a global search at 4a resolution, but resultant handness is still opposite, which raised my concern.

Xing

niko
Offline
Joined: 5 May 2009
There is no flip option in

There is no flip option in the current version of Frealign. In older versions (several years ago) there was a flip option for special cases where a particle had a quasi symmetry. Using the flip option one could test two different possible alignments and try to align particles that were misaligned to the similar quasi symmetrical orientation.

In your case, you will have to change the angles in the parameter file. For example, you could change the last angle (PHI) by replacing it with 360-PHI. This should change the handedness of your reconstruction.

rkman2
Offline
Joined: 18 Nov 2014
Change in convention?

Hi Niko,

Replacing the last angle PHI by 360-PHI (Frealign 9.11) destroys the image reconstruction -output is a sphere for an icosahedral virus. Has the convention changed? If not, is it OK if there are PHI angles greater than 360 given that some of the input values have negative floats.

Happy New Year.

niko
Offline
Joined: 5 May 2009
The formula I provided above

The formula I provided above might have been wrong. I looked through my notes and found another one:

360 - psi, 180 + theta, phi

If this also does not work you may have to work this out yourself (please share here if you do).

xzhang2005
Offline
Joined: 21 Nov 2012
Thanks, Niko. Another related

Thanks, Niko.
Another related question is: How does frealign pickup a consistent handness for all particles during alignment/refinement? since each particle images is aligned/refined independently.

Xing

Axel
Offline
Joined: 20 May 2009
Since the data are all

Since the data are all refined against the same reference, which is handed, they should in theory all have the same handedness during refinement.
I don't know if this properly addresses the question.

Axel

xzhang2005
Offline
Joined: 21 Nov 2012
Thanks, Axel, but this

Thanks, Axel, but this doesn't address my concern. In my case, the reference model has right-handness, but frealign chose left-handness during alignment/refinement, and thus the new reconstruction is left-handed. So my concern is that would frealign choose different handnesses for different input particles? because alignment of each particle image is completely independent from each other.

Xing

niko
Offline
Joined: 5 May 2009
A reference model with a

A reference model with a clear hand will lead to a unique alignment that should reproduce the hand in a reconstruction. The situation you are describing should not occur.

xzhang2005
Offline
Joined: 21 Nov 2012
In my case, the model from

In my case, the model from dataset # is around 3.1a resolution, and I filtered it back to 4a as the referecne model to align/refine the datset #2 binned 2x (nyquist limt=4a). So, I think there should be clear hand info in the 3D model and I did a global search at 4a resolution, but resultant handness is still opposite, which raised my concern.

Xing

niko
Offline
Joined: 5 May 2009
It sounds like you do not

It sounds like you do not have a clear (i.e. visible at low resolution) handedness in your structure. This means that in your initial global search half of your particles align according to one handedness and the other half according to the opposite hand. In your later refinement and global search at higher resolution (4 A) half of your particles will flip their alignment to be consistent with the other half. Which half of the data will "win" this competition is more or less random if you do not have a clear hand at low resolution. If you want to avoid this problem you need to do your initial global search at higher resolution where the handedness of your structure becomes more obvious.

xzhang2005
Offline
Joined: 21 Nov 2012
Does frealign have a

Does frealign have a mechanism to ensure all particles choose the same handness? as my understanding, there is not communication between alignment/refinement of different particles. So, it is still very possible that many particles might be assigned an opposite handness from rest of the particles.

Xing