Memory use, refinement.db, and layer lines

[Patricia Campbell]

Dear Patricia,

Thanks for your questions.
> Hello
>
> I have 4 questions. I tried to get on the spring google group but it doesnt seem to be working.  Version output is Emspring-0.84.1470, EMAN 2.1 alpha2, Python 2.7.2. I have really been enjoying working with Spring. I have been having some great results but I have been running into some roadblocks.
>
>       1.      I have been having some problems with memory allocation during s3drefine runs. My stack is 9.2 GB in size. I attached how the memory was used for the s3drefine run. On that run I used the qsub_SCRcom.txt file. The out-of-memory daemon killed portions of the process to keep the node from crashing. Is there a guide for how memory is used during the run to avoid crashing? I changed it to the qsub_345start-2.txt file. Both setups finished but I just want to make sure that I am not killing portions of the job.
You will have to reserve sufficient memory for the submitted process. This does not depend on the stack size but rather the reconstruction size, which in turn depends on the width of your assembly.
>       2.      I see that someone has asked previously regarding the "too few segments" error. I have been trying to continue refinements by including a previous refinement run with other parameters and had the same error. The file is s3drefine_100-S-PL-PC.txt .
>       ?       Are the orientations from the previous run held in the refinement.db file?
>       ?       How can I avoid this? It seems that I have to either start the refinement without a previous refinement.db file or start all over from the beginning with a collection of parameters with no room for alteration.
>       3.      I wanted to use the layer lines correlation feature in s3drefine. When I turned (I only changed False to True) the option on I got the following error:
>       ?       ArgumentError: Only '=', '!=', 'is_()', 'isnot()' operators can be used with None/True/False
>       ?       how can I avoid this error?
>
Could you send the file via the Google group?

Thanks.

Best wishes,


Carsten
> If I can have a way to get on google groups I can send the aforementioned files. Thank you so much for your time and consideration!
>
> Patricia Campbell
> Graduate Student, James Conway Lab
> Dept. Structural Biology, University of Pittsburgh

[Carsten Sachse]

Dear Patricia,

On Monday, 10 April 2017 16:41:53 UTC+2, Patricia Campbell wrote:

Dear Patricia,

Thanks for your questions.
> Hello
>
> I have 4 questions. I tried to get on the spring google group but it doesnt seem to be working.  Version output is Emspring-0.84.1470, EMAN 2.1 alpha2, Python 2.7.2. I have really been enjoying working with Spring. I have been having some great results but I have been running into some roadblocks.
>
>       1.      I have been having some problems with memory allocation during s3drefine runs. My stack is 9.2 GB in size. I attached how the memory was used for the s3drefine run. On that run I used the qsub_SCRcom.txt file. The out-of-memory daemon killed portions of the process to keep the node from crashing. Is there a guide for how memory is used during the run to avoid crashing? I changed it to the qsub_345start-2.txt file. Both setups finished but I just want to make sure that I am not killing portions of the job.
You will have to reserve sufficient memory for the submitted process. This does not depend on the stack size but rather the reconstruction size, which in turn depends on the width of your assembly.
>       2.      I see that someone has asked previously regarding the "too few segments" error. I have been trying to continue refinements by including a previous refinement run with other parameters and had the same error. The file is s3drefine_100-S-PL-PC.txt .
>       ?       Are the orientations from the previous run held in the refinement.db file?

>       ?       How can I avoid this? It seems that I have to either start the refinement without a previous refinement.db file or start all over from the beginning with a collection of parameters with no room for alteration.

Yes. In case you are right. The previous refinement.db holds all angles. In case you are doing a restrained angular refinement, it will be based on previous refinement.

Taking a closer look at your refinement parameter file, 'too few segments' error may occur that you simply excluded too many. In your case, you specified 2 exclusion criteria, probably you removed too many. For example by:

Persistence class option                 = True

Persistence class length in Angstrom     = 84

Class occupancy threshold                = 0.5

You only consider members of class 84. Is this intentional. There may not be many segments left. Here you can include also list of classes.

In case you would like to better estimate how many segments will be left, you can simply run SEGMENTPLOT and view the segments that make it through your selection procedure.

>       3.      I wanted to use the layer lines correlation feature in s3drefine. When I turned (I only changed False to True) the option on I got the following error:

You will have to predefine that correlation by running them through SEGMENTEXAM. In SEGMENTEXAM you define the area to correlate in Fourier space. SEGMENTEXAM will write out a spring.db. Please continue your refinement with that spring.db as it contains the required correlation values. 

Relevant parameters in SEGMENTEXAM are:

Compute layer-line correlation option    = False

Layer-line region in 1/Angstrom          = (0.03, 0.035)

If you have followed this sequence, the error should not be raised.

I hope this helps.

Best wishes,

Carsten