If you are curious to see a list of the exact refinement parameters used, browse to the 0_refine_parms.json file in the refinement directory. You can use 'Info' in the file browser or just read the file directly (.json files are plain text)
Randomizing the Fourier phases of refine_06/scaled_model.hdf at resolutions higher than 12.0 Å. If the final achieved resolution is not at least ~10.8 Å, then the gold standard resolution assessment is not valid, and you need to re-refine, starting with a lower resolution target.
Input particles are from sets/all-bad1__ctf_flip_lp5.lst
Several different methods can be used for final amplitude correction in cryoEM. The most accurate of these is to take the final structure and filter it so its 1-D power spectrum matches a known 1-D power spectrum from X-ray solution scattering or other source. This 'ideally filtered' structure is then low-pass filtered based on the same FSC curve used to measure resolution between the independent even/odd models used in the Gold Standard resolution criterion. This is the normal method EMAN2 will use for final corrections. Other choices are possible as well, however. EMAN2 generates an experimental structure factor directly from your data as part of the CTF fitting procedure. The low resolution portion is drawn directly from your data and the high resolution portion (generally past 14-18 A) is an empirical structure factor for proteins in general with a mix of alpha and beta components. This works quite well for most proteins, however is not completely appropriate for hybrid molecules with nucleotides (like ribosomes), but will generally still produce reasonable structures. Note that this is just a standard linear filter, which is not really changing the 3-D structure at all, it's really just changing how that structure is portrayed.
The structure factor produced when CTF fitting is stored in strucfac.txt in the project directory. If you replace this file with a structure factor from some other source, it will be used to process all subsequent particles.
You are using the --tophat=global option, which modifies the final filter applied to 3-D maps. The default behavior is to apply a final Wiener filter based on the FSC curve used to compute the resolution. This Wiener filter gives a map which in theory reduces noise and filters the map to get as close as possible to what you should be able to see at the specified resolution. However, this means that some features, such as sidechains and the pitch of alpha-helices may be somewhat smoothed out. The alternative is a 'tophat' or 'sharp' filter imposed at the cutoff (0.143) FSC value, and is what Relion imposes on its final maps. This produces maps which look prettier, with more apparent side-chains at high resolution, but runs the risk that some of these features may be artifacts of the sharp filter.
Your desired resolution is beyond 3/4 Nyquist. Regardless, we will set --classiter to 1 initially. Leaving this above 0 will help avoid noise bias, but it may be reduced to zero if convergence seems to have been achieved.
The resolution you are requesting is beyond 2/3 Nyquist. This is normally not recommended, as it represents insufficient sampling to give a good representation of your reconstructed map, and resolution can be difficult to accurately assess. The reconstruction will proceed, but generally speaking your A/pix should be less than 1/3 the targeted resolution. Based on your requested resolution and box-size, modified by --speed, I will use an angular sampling of 5.00 deg. For details, please see http://blake.bcm.edu/emanwiki/EMAN2/AngStep
As particles get translated during alignment, the total amount of noise present in the aligned particle can change significantly. While this isn't a very large effect for similarity metrics like fsc, it can cause a bias in the reconstruction. Similarly if we use a different mask derived for each projection to combat this problem (as with the zeromask=1 option in some comparators, then each projection masks out a different fraction of the image causing some orientations to be preferred. To combat both effects, I will compute a single aggregate mask from all of the projections in each iteration, and use it as --mask refine_06/simmask.hdf. The mask is autogenerated and overwritten after each iteration. The only way to completely disable this behavior is to specify --simmask yourself with a file containing all 1.0 pixels.
The plot below shows the FSC computed between iterations for both even and odd particle subsets. It is not a measure of resolution, but is a measure of how much the individual even/odd maps are changing from one iteration to the next. In a perfect world, this plot would eventually be 1.0 indicating no change from one iteration to the next after the refinement converges. In reality we normally expect only a pseudoconvergence, where the curves approach a final shape but do not actually become 1.0. This plot is automatically updated after each iteration.
e2refine_easy computes a true gold-standard resolution as part of its stanard processing. This curve is also used as a basis for filtering the maps. Resolution can be gagued as the point at which this curve falls below a value of 0.143. Note that when computing FSCs in other situations, for example, when comparing the final map produced with all of the particle data to a higher resolution crystal structure, the more stringent 0.5 (actually 0.4) criterion must be used. Don't overinterpret these plots. The FSC plots themselves contain some noise, so there is some uncertainty in any resolution value.
Also note that these curves are highly dependent on the mask used prior to FSC computation. See the next plot for more information on this. This iteration comparison plot uses a relatively tight mask designed to be somewhat comparable to that used in Relion (though Relion is still more aggressive in masking).
These resolution curves should start at ~1 at low resolution, fall smoothly to ~0, then remain at roughly zero. The vertical range of an FSC plot is -1 to 1. Some negative oscillations after reaching zero are normal. If the FSC curve falls towards zero, but then rises again at high resolution this is an indication of an artifact, and the resolution values may not be trustworthy. Possible sources include, but are not limited to: mask too tight, box size too small, reconstruction artifacts, ... If you observe an "unhealthy" FSC curve here, please check the next plot, and see if the curve with a looser mask appears "healthier".
As mentioned in the previous section, masking can have a significant impact on cited resolution. For a while users who followed Relion's suggested procedure would find resolution values slightly better than those produced by EMAN. This was due to Relion's use of a very tight mask, often extending into the core of the structure. EMAN2.1 now generates 3 FSC curves automatically: 1) completely unmasked, 2) a loose, "conservative" mask and 3) a tighter mask which should be somewhat similar to Relion 1.3 using its suggested parameters (though EMAN is still slightly more conservative). The plot below compares these three FSC curves for the last completed iteration. The relevant mask files are also stored in the refinement folder as mask.hdf and mask_tight.hdf. Note: if visualizing 3-D masks, it is generally a good idea to display as 2-D slices (single image view in browser) rather than using isosurfaces, so any internal features of the mask can be readily observed.
If you see an 'unheathly' looking unmasked curve (rising at high resolution, etc.), particularly if the masked curves look healthy, this can indicate that your box-size is too small, and you are getting Fourier artifacts at the edge. It could also indicate some other problem, so if re-extracting your boxed particles with a somewhat larger box doesn't fix the problem, please contact us, and we will help you to debug the problem.
If your goal is to compare EMAN2.1 results to Relion, the only way to reliably accomplish this is to extract the unmasked even and odd gold-standard maps from both EMAN and RELION, mask them with exactly the same mask, then compute the FSC. For visual comparison of the final gold-standard maps from both packages, you must insure that they are both identically filtered. This can be done easily by matching the 1-D power spectrum of one of the maps to the other. For example e2proc3d.py map_eman.hdf map_eman_cmp.hdf --process filter.matchto:to=map_relion.mrc
Auto amplitude correction using mode:strucfac in this iteration
Iteration 1: Resolution = 8.5 Å (gold standard refinement with tight mask, FSC @0.143)
Iteration 1: Resolution with different masks = 8.5, 8.7, 11.8 Å
Auto amplitude correction using mode:flatten in this iteration
Iteration 2: Resolution = 8.7 Å (gold standard refinement with tight mask, FSC @0.143)
Iteration 2: Resolution with different masks = 8.7, 11.4, 11.7 Å
Auto amplitude correction using mode:flatten in this iteration
Iteration 3: Resolution = 9.6 Å (gold standard refinement with tight mask, FSC @0.143)
Iteration 3: Resolution with different masks = 9.6, 11.3, 11.4 Å
Auto amplitude correction using mode:flatten in this iteration
Iteration 4: Resolution = 10.7 Å (gold standard refinement with tight mask, FSC @0.143)
Iteration 4: Resolution with different masks = 10.7, 10.7, 11.1 Å
Congratulations, your refinement is complete, and you have a gold standard resolution of 10.7 Å. Note that there is always some variability in these determined values based on masking of the map. If the map is masked too tightly, the FSC curve will not remain near zero at high resolution and will rise again. If you see an FSC curve that falls to zero then rises again, it is an indication that something may have gone wrong with the masking. In general, EMAN2.1 is very conservative in its masking to try and avoid these problems. This means that the resolution may be as much as 10% better with somewhat tighter masking, without getting into artifact territory.
If you wish to continue this refinement to further improve resolution, the most efficient approach is to use the --startfrom refine_XX option rather than specifying --input and --model. When you use --startfrom, it will not re-randomize the phases. Since you have already achieved sufficient resolution to validate the gold-standard approach, continuing to extend this resolution is valid, and more efficient.
Here are some useful output files to look at:
Sat Aug 14 22:40:30 2021: e2proc3d.py refine_05/threed_02.hdf refine_06/scaled_model.hdf --clip=132,132,132 --scale=1.57143
Sat Aug 14 22:40:31 2021: e2proc3d.py refine_06/scaled_model.hdf refine_06/threed_00_even.hdf --process=filter.lowpass.randomphase:cutoff_freq=0.08333333333333333 --apix=2.4727272987365723 --compressbits 10
Sat Aug 14 22:40:32 2021: e2proc3d.py refine_06/scaled_model.hdf refine_06/threed_00_odd.hdf --process=filter.lowpass.randomphase:cutoff_freq=0.08333333333333333 --apix=2.4727272987365723 --compressbits 10
Sat Aug 14 22:40:32 2021: e2proc3d.py refine_06/threed_00_even.hdf refine_06/fsc_unmasked_00.txt --calcfsc refine_06/threed_00_odd.hdf
*** Changing classiter from 1 to 1 ***
* Generating 2-D projections of even/odd 3-D maps
Sat Aug 14 22:40:33 2021: e2project3d.py refine_06/threed_00_even.hdf --outfile refine_06/projections_01_even.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
Sat Aug 14 22:40:55 2021: e2project3d.py refine_06/threed_00_odd.hdf --outfile refine_06/projections_01_odd.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
* Computing similarity of each particle to the set of projections using a hierarchical scheme. This will be the basis for classification.
Sat Aug 14 22:41:12 2021: e2simmx2stage.py refine_06/projections_01_even.hdf sets/all-bad1__ctf_flip_lp5_even.lst refine_06/simmx_01_even.hdf refine_06/proj_simmx_01_even.hdf refine_06/proj_stg1_01_even.hdf refine_06/simmx_stg1_01_even.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
Sun Aug 15 00:48:40 2021: e2simmx2stage.py refine_06/projections_01_odd.hdf sets/all-bad1__ctf_flip_lp5_odd.lst refine_06/simmx_01_odd.hdf refine_06/proj_simmx_01_odd.hdf refine_06/proj_stg1_01_odd.hdf refine_06/simmx_stg1_01_odd.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
* Based on the similarity values, put each particle in to 1 or more classes (depending on --sep)
Sun Aug 15 02:50:16 2021: e2classify.py refine_06/simmx_01_even.hdf refine_06/classmx_01_even.hdf -f --sep 1
Sun Aug 15 02:50:48 2021: e2classify.py refine_06/simmx_01_odd.hdf refine_06/classmx_01_odd.hdf -f --sep 1
* Iteratively align and average all of the particles within each class, discarding the worst fraction
Sun Aug 15 02:51:22 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_even.lst --classmx refine_06/classmx_01_even.hdf --decayedge --storebad --output refine_06/classes_01_even.hdf --ref refine_06/projections_01_even.hdf --iter 1 -f --resultmx refine_06/cls_result_01_even.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
Sun Aug 15 02:56:09 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_odd.lst --classmx refine_06/classmx_01_odd.hdf --decayedge --storebad --output refine_06/classes_01_odd.hdf --ref refine_06/projections_01_odd.hdf --iter 1 -f --resultmx refine_06/cls_result_01_odd.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
* Using the known orientations, reconstruct the even/odd 3-D maps from the even/odd 2-D class-averages.
Sun Aug 15 03:00:54 2021: e2make3dpar.py --input refine_06/classes_01_even.hdf --sym d2 --output refine_06/threed_01_even.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10
Sun Aug 15 03:03:42 2021: e2make3dpar.py --input refine_06/classes_01_odd.hdf --sym d2 --output refine_06/threed_01_odd.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10
* Finally, determine the resolution, filter and mask the even/odd maps, and then produce the final 3-D map for this iteration. Note that the next iteration is seeded with the individual even/odd maps, not the final average.
Sun Aug 15 03:06:35 2021: e2refine_postprocess.py --even refine_06/threed_01_even.hdf --odd refine_06/threed_01_odd.hdf --output refine_06/threed_01.hdf --automaskexpand -1 --align --mass 400.0 --iter 1 --setsf strucfac.txt --tophat=global --sym d2 --restarget 6.0 --underfilter --ampcorrect strucfac --compressbits 10 --threads 4
Sun Aug 15 03:06:50 2021: e2proc3d.py refine_06/threed_01_even.hdf refine_06/converge_even_00_01.txt --calcfsc refine_06/threed_00_even.hdf
Sun Aug 15 03:06:51 2021: e2proc3d.py refine_06/threed_01_odd.hdf refine_06/converge_odd_00_01.txt --calcfsc refine_06/threed_00_odd.hdf
*** Changing classiter from 1 to 0 ***
* Generating 2-D projections of even/odd 3-D maps
Sun Aug 15 03:06:53 2021: e2project3d.py refine_06/threed_01_even.hdf --outfile refine_06/projections_02_even.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
Sun Aug 15 03:07:10 2021: e2project3d.py refine_06/threed_01_odd.hdf --outfile refine_06/projections_02_odd.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
* Computing similarity of each particle to the set of projections using a hierarchical scheme. This will be the basis for classification.
Sun Aug 15 03:07:27 2021: e2simmx2stage.py refine_06/projections_02_even.hdf sets/all-bad1__ctf_flip_lp5_even.lst refine_06/simmx_02_even.hdf refine_06/proj_simmx_02_even.hdf refine_06/proj_stg1_02_even.hdf refine_06/simmx_stg1_02_even.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
Sun Aug 15 05:06:42 2021: e2simmx2stage.py refine_06/projections_02_odd.hdf sets/all-bad1__ctf_flip_lp5_odd.lst refine_06/simmx_02_odd.hdf refine_06/proj_simmx_02_odd.hdf refine_06/proj_stg1_02_odd.hdf refine_06/simmx_stg1_02_odd.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
* Based on the similarity values, put each particle in to 1 or more classes (depending on --sep)
Sun Aug 15 07:07:08 2021: e2classify.py refine_06/simmx_02_even.hdf refine_06/classmx_02_even.hdf -f --sep 1
Sun Aug 15 07:07:41 2021: e2classify.py refine_06/simmx_02_odd.hdf refine_06/classmx_02_odd.hdf -f --sep 1
* Iteratively align and average all of the particles within each class, discarding the worst fraction
Sun Aug 15 07:08:14 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_even.lst --classmx refine_06/classmx_02_even.hdf --decayedge --storebad --output refine_06/classes_02_even.hdf --ref refine_06/projections_02_even.hdf --iter 0 -f --resultmx refine_06/cls_result_02_even.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
Sun Aug 15 07:11:14 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_odd.lst --classmx refine_06/classmx_02_odd.hdf --decayedge --storebad --output refine_06/classes_02_odd.hdf --ref refine_06/projections_02_odd.hdf --iter 0 -f --resultmx refine_06/cls_result_02_odd.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
* Using the known orientations, reconstruct the even/odd 3-D maps from the even/odd 2-D class-averages.
Sun Aug 15 07:14:13 2021: e2make3dpar.py --input refine_06/classes_02_even.hdf --sym d2 --output refine_06/threed_02_even.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10 --itermask refine_06/mask.hdf
Sun Aug 15 07:17:01 2021: e2make3dpar.py --input refine_06/classes_02_odd.hdf --sym d2 --output refine_06/threed_02_odd.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10 --itermask refine_06/mask.hdf
* Finally, determine the resolution, filter and mask the even/odd maps, and then produce the final 3-D map for this iteration. Note that the next iteration is seeded with the individual even/odd maps, not the final average.
Sun Aug 15 07:19:53 2021: e2refine_postprocess.py --even refine_06/threed_02_even.hdf --odd refine_06/threed_02_odd.hdf --output refine_06/threed_02.hdf --automaskexpand -1 --align --mass 400.0 --iter 2 --setsf strucfac.txt --tophat=global --sym d2 --restarget 6.0 --underfilter --ampcorrect flatten --compressbits 10 --threads 4
Sun Aug 15 07:20:08 2021: e2proc3d.py refine_06/threed_02_even.hdf refine_06/converge_even_01_02.txt --calcfsc refine_06/threed_01_even.hdf
Sun Aug 15 07:20:08 2021: e2proc3d.py refine_06/threed_02_odd.hdf refine_06/converge_odd_01_02.txt --calcfsc refine_06/threed_01_odd.hdf
* Generating 2-D projections of even/odd 3-D maps
Sun Aug 15 07:20:10 2021: e2project3d.py refine_06/threed_02_even.hdf --outfile refine_06/projections_03_even.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
Sun Aug 15 07:20:27 2021: e2project3d.py refine_06/threed_02_odd.hdf --outfile refine_06/projections_03_odd.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
* Computing similarity of each particle to the set of projections using a hierarchical scheme. This will be the basis for classification.
Sun Aug 15 07:20:44 2021: e2simmx2stage.py refine_06/projections_03_even.hdf sets/all-bad1__ctf_flip_lp5_even.lst refine_06/simmx_03_even.hdf refine_06/proj_simmx_03_even.hdf refine_06/proj_stg1_03_even.hdf refine_06/simmx_stg1_03_even.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
Sun Aug 15 09:20:29 2021: e2simmx2stage.py refine_06/projections_03_odd.hdf sets/all-bad1__ctf_flip_lp5_odd.lst refine_06/simmx_03_odd.hdf refine_06/proj_simmx_03_odd.hdf refine_06/proj_stg1_03_odd.hdf refine_06/simmx_stg1_03_odd.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
* Based on the similarity values, put each particle in to 1 or more classes (depending on --sep)
Sun Aug 15 11:30:35 2021: e2classify.py refine_06/simmx_03_even.hdf refine_06/classmx_03_even.hdf -f --sep 1
Sun Aug 15 11:31:17 2021: e2classify.py refine_06/simmx_03_odd.hdf refine_06/classmx_03_odd.hdf -f --sep 1
* Iteratively align and average all of the particles within each class, discarding the worst fraction
Sun Aug 15 11:32:00 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_even.lst --classmx refine_06/classmx_03_even.hdf --decayedge --storebad --output refine_06/classes_03_even.hdf --ref refine_06/projections_03_even.hdf --iter 0 -f --resultmx refine_06/cls_result_03_even.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
Sun Aug 15 11:35:17 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_odd.lst --classmx refine_06/classmx_03_odd.hdf --decayedge --storebad --output refine_06/classes_03_odd.hdf --ref refine_06/projections_03_odd.hdf --iter 0 -f --resultmx refine_06/cls_result_03_odd.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
* Using the known orientations, reconstruct the even/odd 3-D maps from the even/odd 2-D class-averages.
Sun Aug 15 11:38:26 2021: e2make3dpar.py --input refine_06/classes_03_even.hdf --sym d2 --output refine_06/threed_03_even.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10 --itermask refine_06/mask.hdf
Sun Aug 15 11:41:28 2021: e2make3dpar.py --input refine_06/classes_03_odd.hdf --sym d2 --output refine_06/threed_03_odd.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10 --itermask refine_06/mask.hdf
* Finally, determine the resolution, filter and mask the even/odd maps, and then produce the final 3-D map for this iteration. Note that the next iteration is seeded with the individual even/odd maps, not the final average.
Sun Aug 15 11:44:31 2021: e2refine_postprocess.py --even refine_06/threed_03_even.hdf --odd refine_06/threed_03_odd.hdf --output refine_06/threed_03.hdf --automaskexpand -1 --align --mass 400.0 --iter 3 --setsf strucfac.txt --tophat=global --sym d2 --restarget 6.0 --underfilter --ampcorrect flatten --compressbits 10 --threads 4
Sun Aug 15 11:44:46 2021: e2proc3d.py refine_06/threed_03_even.hdf refine_06/converge_even_02_03.txt --calcfsc refine_06/threed_02_even.hdf
Sun Aug 15 11:44:46 2021: e2proc3d.py refine_06/threed_03_odd.hdf refine_06/converge_odd_02_03.txt --calcfsc refine_06/threed_02_odd.hdf
* Generating 2-D projections of even/odd 3-D maps
Sun Aug 15 11:44:49 2021: e2project3d.py refine_06/threed_03_even.hdf --outfile refine_06/projections_04_even.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
Sun Aug 15 11:45:06 2021: e2project3d.py refine_06/threed_03_odd.hdf --outfile refine_06/projections_04_odd.hdf -f --projector standard --orientgen eman:delta=4.99944:inc_mirror=0:perturb=0 --sym d2 --parallel thread:4 --compressbits 10
* Computing similarity of each particle to the set of projections using a hierarchical scheme. This will be the basis for classification.
Sun Aug 15 11:45:23 2021: e2simmx2stage.py refine_06/projections_04_even.hdf sets/all-bad1__ctf_flip_lp5_even.lst refine_06/simmx_04_even.hdf refine_06/proj_simmx_04_even.hdf refine_06/proj_stg1_04_even.hdf refine_06/simmx_stg1_04_even.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
Sun Aug 15 13:56:41 2021: e2simmx2stage.py refine_06/projections_04_odd.hdf sets/all-bad1__ctf_flip_lp5_odd.lst refine_06/simmx_04_odd.hdf refine_06/proj_simmx_04_odd.hdf refine_06/proj_stg1_04_odd.hdf refine_06/simmx_stg1_04_odd.hdf --saveali --cmp frc:snrweight=1:maxres=7.0 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:zeromask=1:snrweight=1:minres=80:maxres=6.0 --shrinks1 2 --mask refine_06/simmask.hdf --parallel thread:4
* Based on the similarity values, put each particle in to 1 or more classes (depending on --sep)
Sun Aug 15 16:05:29 2021: e2classify.py refine_06/simmx_04_even.hdf refine_06/classmx_04_even.hdf -f --sep 1
Sun Aug 15 16:06:03 2021: e2classify.py refine_06/simmx_04_odd.hdf refine_06/classmx_04_odd.hdf -f --sep 1
* Iteratively align and average all of the particles within each class, discarding the worst fraction
Sun Aug 15 16:06:37 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_even.lst --classmx refine_06/classmx_04_even.hdf --decayedge --storebad --output refine_06/classes_04_even.hdf --ref refine_06/projections_04_even.hdf --iter 0 -f --resultmx refine_06/cls_result_04_even.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
Sun Aug 15 16:09:46 2021: e2classaverage.py --input sets/all-bad1__ctf_flip_lp5_odd.lst --classmx refine_06/classmx_04_odd.hdf --decayedge --storebad --output refine_06/classes_04_odd.hdf --ref refine_06/projections_04_odd.hdf --iter 0 -f --resultmx refine_06/cls_result_04_odd.hdf --normproc normalize.edgemean --averager ctf.weight --keep 0.9 --cmp frc:snrweight=1 --align rotate_translate_flip:usebispec=1 --aligncmp ccc --ralign refine --raligncmp frc:snrweight=1:zeromask=1:minres=80:maxres=6.0 --compressbits 10 --parallel thread:4
* Using the known orientations, reconstruct the even/odd 3-D maps from the even/odd 2-D class-averages.
Sun Aug 15 16:12:51 2021: e2make3dpar.py --input refine_06/classes_04_even.hdf --sym d2 --output refine_06/threed_04_even.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10 --itermask refine_06/mask.hdf
Sun Aug 15 16:15:57 2021: e2make3dpar.py --input refine_06/classes_04_odd.hdf --sym d2 --output refine_06/threed_04_odd.hdf --keep 0.8 --apix 2.4727272987365723 --pad 180 --iterative --threads 4 --compressbits 10 --itermask refine_06/mask.hdf
* Finally, determine the resolution, filter and mask the even/odd maps, and then produce the final 3-D map for this iteration. Note that the next iteration is seeded with the individual even/odd maps, not the final average.
Sun Aug 15 16:18:58 2021: e2refine_postprocess.py --even refine_06/threed_04_even.hdf --odd refine_06/threed_04_odd.hdf --output refine_06/threed_04.hdf --automaskexpand -1 --align --mass 400.0 --iter 4 --setsf strucfac.txt --tophat=global --sym d2 --restarget 6.0 --underfilter --ampcorrect flatten --compressbits 10 --threads 4
Sun Aug 15 16:19:13 2021: e2proc3d.py refine_06/threed_04_even.hdf refine_06/converge_even_03_04.txt --calcfsc refine_06/threed_03_even.hdf
Sun Aug 15 16:19:14 2021: e2proc3d.py refine_06/threed_04_odd.hdf refine_06/converge_odd_03_04.txt --calcfsc refine_06/threed_03_odd.hdf