 |
 |
This makes sense. As per my and Juhas suggestions, local alignments and/or very high healpix setting is necessary for large and high-symmetry objects to mimic the high-sampling behavior of auto-refinement, and reach comparable resolutions. Still, the division of data over many object might prove detrimental to the final resolution, more than any gain you might get from classifying, so resolution is in no way guaranteed to improve or even be the same. /B On 08/07/2018 10:13 AM, Dieter Blaas wrote: > > Dear Björn, > > thanks for the hint! I am sorry for the confusion: healpix is not > changing over the classification run! I only noticed that my Class3D > run (with alignment) converged at a very low resolution (8.3 A) > despite using a run_data.star file from 3D auto-refine that had > converged at much higher resolution (3.8 A). I am now running the > classification without alignment and this yields classes with much > better resolution (right now at 4.45 A) but still running! > > bw, Dieter > > ------------------------------------------------------------------------ > Dieter Blaas, > Max F. Perutz Laboratories > Medical University of Vienna, > Inst. Med. Biochem., Vienna Biocenter (VBC), > Dr. Bohr Gasse 9/3, > A-1030 Vienna, Austria, > Tel: 0043 1 4277 61630, > Fax: 0043 1 4277 9616, > e-mail:[log in to unmask] > ------------------------------------------------------------------------ > Am 07.08.2018 um 09:47 schrieb Björn Forsberg: >> Not sure what you mean... how is the healpix increasing in a run >> without alignment? >> >> Also, classifications never increase the sampling dynamically, so >> what you are observing is pretty expected. For high-res >> classifications, I'd recommend starting with local search and giving >> an Autorefine data.star as input. >> >> Cheers, >> >> Björn >> ------------------------------------------------------------------------ >> *Från:* Collaborative Computational Project in Electron >> cryo-Microscopy <[log in to unmask]> för Dieter Blaas >> <[log in to unmask]> >> *Skickat:* den 4 augusti 2018 22:56:57 >> *Till:* [log in to unmask] >> *Ämne:* [ccpem] relion-3.0-beta Class3D converges at very low resolution >> >> Hi Sjors and all, >> >> using relion-3.0-beta2 I refined a virus to 3.8 A (~ 60,000 >> particles) by using symmetry. Trying to push the resolution further I >> run a Class3D with 3 classes but it never ever attained a resolution >> better than 8.39 A and the Healpixorder remained 3 and never >> increased. What might be wrong? A similar run _without alignment_ >> resulted in a resolution of 4.79 A for the best class and the >> Healpixorder increased from 2 to 7 during the run. >> >> Here is the command: `which relion_refine_mpi` --o Class3D/job108/run >> --i Refine3D/job102/better.star --ref >> PostProcess/job103/postprocess_masked.mrc --ini_high 60 >> --dont_combine_weights_via_disc --pool 3 --pad 2 --ctf >> --ctf_corrected_ref --iter 200 --tau2_fudge 4 --particle_diameter 400 >> --K 3 --flatten_solvent --oversampling 1 --healpix_order 3 >> --sigma_ang 1.66667 --offset_range 5 --offset_step 2 --sym I2 --norm >> --scale --j 2 --gpu "" >> >> I do not know where the parameter "healpix_order 3" comes from as I >> did not define it. >> >> >> Thanks for hints, bw Dieter >> >> >> ------------------------------------------------------------------------ >> Dieter Blaas, >> Max F. Perutz Laboratories >> Medical University of Vienna, >> Inst. Med. Biochem., Vienna Biocenter (VBC), >> Dr. Bohr Gasse 9/3, >> A-1030 Vienna, Austria, >> Tel: 0043 1 4277 61630, >> Fax: 0043 1 4277 9616, >> e-mail:[log in to unmask] >> ------------------------------------------------------------------------ >> >> ------------------------------------------------------------------------ >> >> To unsubscribe from the CCPEM list, click the following link: >> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1 >> > ######################################################################## To unsubscribe from the CCPEM list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1