I have been running numerous RBMC jobs recently and have observed that the ‘remove duplicated particles’ job does not consistently handle all duplicated particles identified by the RBMC job. I tried 50, 70, 120, and 150 Å distance for removal. Is it feasible to incorporate an automatic process within the RBMC job to manage duplicated particles? This issue has already caused my job to halt on three occasions.
[CPU: 25.0 MB Avail: 250.57 GB]
====== Job process terminated abnormally.
[CPU: 212.0 MB Avail: 250.46 GB]
DIE: particle pairwise distance matrix is singular, check for duplicate particles (sgetrf failed)
We’re aware that this has been a pain point for many users, I apologize for the inconvenience. In an upcoming CryoSPARC release, we’ll be adjusting this behaviour so that the micrograph will be skipped rather than the whole job failing. I know that’s only a partial solution, but it should help at least somewhat.
Hi @hsnyder! Thank you for your response! Wouldn’t it be easier for RBMC to implement a process that removes duplicated particles internally per micrograph and retains only the particle with the best statistics from the 3D alignment error metric, for example? While skipping the micrograph is a viable solution, there is a risk of skipping all micrographs if the particles are not properly prepared for RBMC.
Yes, as I said it’s only a partial solution. At some point we will be doing a set of more thorough improvements to RBMC but the timeline for that is further out.