Controller receives disconnect before job was finished: 2DTM

[bencreekm]

Hello,

I have gone through the tutorial successfully for 2DTM, but am trying to implement this function with my own data. I have had this issue with several different images, but am not sure what the error means or where to find more information.

With about 18-19s predicted left in 2DTM this error shows up in the GUI:
Controller received a disconnect from the master before the job was finished..
Error: Controller Disconnected - shutting down job.

There is no corresponding output in the terminal except that it occurs right after:
total number 398855

Do you know what may be occurring? or is there somewhere I can check for more information on the error?
Thanks!

[bHimes]

This is an unfortunately nondescript error message. All it means is that the worker (the process actually running an instance of match_template) has crashed.

I would try three things to narrow it down.

1. Search your image and reference using exactly the same search parameters as the tutorial
2. If it still crashes, using your image, but the reference from the tutorial
3. If that crashes, try using the tutorial image, and your reference.

[bencreekm]

Using the exact same parameters as the tutorial (including point group symmetry), the search completed successfully, but without finding a target likely because my template had a symmetry of C6 and I chose O like the tutorial. Using my template with C1 or C2 point group symmetry yielded the same error as before.

Using the tutorial template, it completed successfully on my image with both C1 and O with a few false positive hits.

So I guess there is something about my template that is a problem. I have tried 2 different pdb entries as templates with no luck. To generate the templates I used the same script that I used to make the 2w0o template following:

1. download the biologic assembly pdb file
2. change the pixel size to match the reported resolution of the structure (3.2A for one and 3.9A for the other)
3. change CS to 2.7 mm as reported (I also tried with 0.001 with no luck)

All other parameters I left the same. I made sure to add the correct pixel size when importing the 3D volume

[bHimes]

your pixel size should match the pixel size of the image you search, not the spatial resolution of the map the model is built from.

Maybe thats it?

Even if you image were a 3.4 Ang/pixel this would be too low res for reliable 2DTM (Nyquist is only 6.8A in that case)

[bencreekm]

Ah okay. I misunderstood that parameter. My image pixel size is 1.08A so that should probably be okay.

I regenerated the two templates I tried previously with the image pixel size and performed the search again, but had the same error.

[bHimes]

What is your run profile for this search, and which card(s) are you running on ?

Could you post a pic of the image you are searching? (you can copy paste a screen shot of either the ctffind or unblur result panel in cisTEM, which displays the image in your reply)

[bencreekm]

My run profile is the gpu_1 that I setup for the tutorial. I am using 1 NVIDIA Quadro RTX 5000
No. Copies: 1
No. Threads Per Copy: 4
Override Total No. Copies: No
Overriden Total No. Copies: 0
Launch Delay (ms): 10

Would it be possible to send you the image by email or in a private message?

[bHimes]

Is it top secret? I prefer to do things in the open ... also, then the time i spend on this, may be useful for others.

[bencreekm]

After consulting with my mentors, we don't feel comfortable sharing the images publicly yet. We are working in a hot field right now. However, once we are ready to go public with the information, we would be happy to upload the images to github if that would be beneficial to others as well

[bHimes]

Thanks for checking @bencreekm, I understand this is a perennial concern in some fields/topics. I'm fortunate to work on boring enough things that I don't often worry too much about being scooped : )

If you are willing to post an image, after you publish, so that there is still a useful archive (hopefully preventing some duplication of work on my part) then I would be okay with working "behind the curtains" for now. I've updated my profile so that my email address is visible on my github page.