confusing alignment result

Hi,

I'm using unblur (by itself, not with summovie) to output aligned averages. I have encountered a dataset that seems particularly problematic where a lot of the stacks seem like they get a couple of misaligned frames (example attached), but when I check the shifts file, all the shifts seem reasonable... Is this artifact not due to frame misalignment?

# Unblur shifts file for input stack : /tmp/unblurumn099wn/input.mrc
# Shifts below are given in Angstroms
# Number of micrographs: 1
# Number of frames per movie: 20
# Pixel size (A): .6650
# 2 lines per micrograph. 1: X-Shift (A); 2: Y-Shift (A)
# -------------------------
# Micrograph 1 of 1
6.799996 6.707618 6.114736 5.221993 3.963431 2.785576 1.942384 1.155619 0.4909202 0.2802958 0.000000 -0.3000240 -0.3874783 -0.6928769 -0.7636621 -0.8978149 -0.8530962 -0.8116279 -1.152586 -0.4990690
-6.309247 -5.847846 -5.374187 -4.677651 -3.820975 -3.044326 -2.405472 -1.743730 -1.081843 -0.5508275 0.000000 0.5969692 1.256582 1.890533 2.351381 2.709984 2.876412 2.902637 3.149904 3.517951

Hi,

The alignment looks good so it is probably not down to the alignment. If you just sum the movie without alignment do you see this artifact? If you look at the individual frames do you see anything weird?

Tim

In reply to by timgrant

No, it is not there. Furthermore I re-ran the image with the exact same options and this time the result looked OK. I've now repeated this same behavior for several other images. Note: the process may have run on a different machine each time.

Does unblur use some initial randomization?

I've found if I rerun these problem images enough times it eventually will produce a normal looking average. Is it possible there is a memory allocation that is assuming zeroed memory?

In reply to by craigyk

Hmm that is indeed weird.

What version of unblur is this?

Are the alignment results the same everytime?

Tim

In reply to by timgrant

Hi Tim,

It is unblur 1.0.2

The runs are not identical, bur generally within 1Å each other. But even if they are within an almost identical range, the averaged sum could have artifacts, or be completely free of them.

No, it is not there. Furthermore I re-ran the image with the exact same options and this time the result looked OK. I've now repeated this same behavior for several other images. Note: the process may have run on a different machine each time.

Does unblur use some initial randomization?