Helical processing with many data sets

Submitted by Unknown on Tue, 2011-08-02 17:34

Forums

Frealign

I am trying to run frealign on helical segments and have a couple of problems with it and am hoping you might be able to help.

Starting from the latest version of frealign_v8_09, I have 77 datasets to combine so have had to increase the value of MAXSET and recompile. I now find that one of the parameters you need for helices is NU (no subunits per segment) and no longer works as it expects an integer. I do not have an integral number of subunits per segment so I need to use a floating point number here (which worked with the already compiled version I downloaded). Is it the case that maybe you have an updated card5.f which can read a real number? And if so, can it manage later on?

The other problem is much more difficult. I am finding that many of the angles calculated for the segments are way out. For myself, I always align the segments vertically then run Ed's IHRSR. This provide the symmetry fairly accurately for input to frealign. If possible it would be wonderful to have an option to restrict the angular search from, say -10 to +10 degrees from the vertical.

niko Tue, 2011-08-02 17:36

The subunits per segment can be set to the nearest integer. This value is only necessary to determine how many times a segment is being reused in the reconstruction step (similar to the n-fold symmetry in a point group). If this is not completely accurate it will not affect the reconstruction. One should avoid setting it to something much smaller or bigger, though.

The angular alignment of segments is actually restrained by the stiffness parameter. However, since I could not assume any particular alignment of a filament (for example vertical, as in your case) the restraint is always trying to keep all segments close to the average in-plane and out-of-plane angles. Also, the restraint is related to how variable the alignment of the segments is. If you already have all the segments nicely aligned and the filament is pretty rigid (e.g. TMV) the variability is very small and segments are more tightly restrained than in the case of a flexible filament (e.g. amyloid fibril).

To help the refinement, I also wrote a small additional program called set_polarity. It reads in a Frealign parameter file and resets the angles of the segments to achieve a consistent polarity. Have a look in the refine.com script in the helical example included with Frealign. I am not sure how you obtain initial angles for your segments but it might be best to do a rough alignment with Spider or Ed's program and then transfer these parameters to Frealign. Then you will be able to use the stiffness parameter to enforce consistent in-plane and out-of-plane angles (higher stiffness numbers give stronger restraints). Using roughly determined parameters is usually always a good idea with Frealign, rather than letting Frealign to the initial search. Frealign's search algorithms are not that efficient. Hope this helps.

Unknown Tue, 2011-08-02 17:39

I have managed to calculate a nice map with -ve stain data to ~20A, turning off the CTF correction despite the inappropriate rotation of some of the segments. However, I cannot get any of the cryo data to succeed either with, or without the CTF correction. I have thousands of these errors and would be grateful if you could throw some light on what this might mean:

 VA04A ENDLESS LOOP SAFETY CATCH: ICNT =         5001
 CC for particle   6598 is -0.05915374  phase residual is 360.00
 **** PROJECTION  6598 ABOVE THRESHOLD, NOT INCLUDED
 values of PSI,THETA,PHI at FMATCH extraction    0.000   90.000    0.000

This only happens when I am trying to combine many datasets together from parameter files created from running frealign on individual stacks (from separate micrographs).