Appion

I see that prepFrealign.py which we retire at 2.2 converts all icos model into (2 3 5) in Frealign refinement but does not do anything special upon upload which would mean that the uploaded euler could not be used to project the initial model database entry into the view of the particle although would be fine with the refined density.

Current prepRefineFrealign.py retains (2 3 5) and (2 5 3) that frealign accepts, but it therefore can not accept (5 3 2) unless converted.

Was it meant to save in database eulers only one icos convention regardless of package ?

Hi Anchi,

Yeah it was not rotating my model correctly, so I rotated it by hand. It was a mess, my frealign jobs would hang forever - probably my install of TORQUE and the agent.py - so I had to manually run all the job files. prepFrealign and uploadFrealign worked great other than the rotation issue. Eulers were converted fine from EMAN to FREALIGN.

I find that to convert between EMAN and FREALIGN some virus models needed a 90 rotation before proc3d icos5fTo2f and others did not.

r17634 enforces the (2 3 5) for Xmip and Frealign during refinement.

Here is what I know:
Frealign I symmetry is the same as Xmipp I1 (2 3 5)
Frealign I2 symmetry is the same as Xmipp I2 (2 5 3)

We have not had a native EMAN icos reconstruction for a while. I may have to pull some old ones out to be sure of its true orientation.

Looks like EMAN (5 3 2) is the same as Xmipp I4 (5 3 2) 5-fold on z, a 2-fold on y, one of the front most 5-fold on xz-plane pointing to positive x.

Confirmed by old eman reconstructions of rotavirus and two uploaded models (afhv and cnv). I did icos5fTo2f followed by rot=0,0,90 All gave the same result of (2 3 5) Viper.

I will add what I know. I've never been able to get the reconstruction stuff working properly here at FSU (I've never had the time to troubleshoot). Anyway, I hacked into the Appion code, so that I don't have to upload reconstructions. My current workflow works with EMAN and Frealign and uses Appion libraries but not database. I use EMAN do get initial icosahedral Eulers (5 fold on Z). Then, use Appion code to convert those to the Frealign convention (2-fold on Z). Then I use the newly generated Frealign param file to make a reconstruction in Frealign. The resulting reconstruction will be in the correct orientation. Then I run a Frealign refinement with runFrealign.py that is in Appion but may only work at FSU.

r17675 and r17659 debug r17632 prepRefineEman.py

r17678 consider default orientation at each conversion point so that the eulers are correct for the default at each point.
EMAN, database: 532
Frealign,Xmipp, particle.txt: 235

The refinement and upload are now correct for both starting orientation between EMAN and Frealign.

The remaining problems:
1. The rotated model has a 0.5 pixel shift EMAN failed the test dataset because of it.
2. A rotation of the reconstructed volume back to the initial model might be needed? Not sure if it causes too much confusion.

Anchi,
Are you leaving this open for your point that "A rotation of the reconstructed volume back to the initial model might be needed?"?