Change Log - StochasticAnalytics/emClarity GitHub Wiki

2019-Apr-15

Further fixes to the auto tilt order generation for the case where some views are to be ignored. T.F. pointed out that the xforms were not being properly modified. This commit addresses this issue. Any local transforms will still be wrong though and I should probably include line to automatically mv any tiltN.local to tiltN.local~ so they aren't used.

2019-Apr-12

Uploaded emClarity 1.2.1

This has some minor fixes for ctf estimate which won't affect everyone, but if you get an error just after the astigmatism fit, this fix is for you.

Also caught a minor bug when specifying the "original" dose symmetric scheme.

2019-Apr-9

Uploaded emClarity 1.2.0 I'm not quite ready to release the version with the speed improvements as I'm still debugging and focusing on maintaining a more stable release model.

The main feature of this release is improved visual interpretation (by eye) of tomograms, I have implemented my take on the deconvolution filter describted by Dimitry Tegunov in his preprint on Warp

To reconstruct the full area for picking regions of interest you can run:

$ emClarity ctf 3d paramN.m tiltN 4

The output will be cache/tiltN_binN_filtered.rec, which you'll need to move to your bin10 folder and edit this or your recScript2.sh to use the proper binning. (Getting rid of these scripts is a high priority, and will happen in the GUI!)

Alternatively, if you just want to visualize all your tomograms at some stage of aligment

set phakePhasePlate=1 in your paramN.m

$ emClarity ctf 3d paramN.m

the output will be tomograms at whatever your aliSamplingRate is.

2019-Apr-1

No April fools here - a patch for latest update is here emClarity_1.1.9.2

2019-Mar-27

The binary for emClarity_1.1.9 is available.

Switch to new mcr '''2019a''' in particular to make running on RTX cards more reliable.
Fix on the Hagen scheme to allow for "grouped" by default, and the original scheme by specifying a negative doseSymmetricIncrement.
Fix a bug that introduced in 1.1.7 that should only affect higher resolution projects. The dose weighting is applied correctly to the projection images prior to reconstruction (most important) however the tilt angles were inverted in the 3D sampling function. This will make a very small difference with dose-symmetric data, but is fixed in any case.
A number of small low-level bug fixes.
This will be the last update before vs. 1.2.0 which will run ~ 2x as fast as current versions! 2019-Mar-1

The new binary for emClarity_1.1.7 is available.

new bugs! I realized the fix for the Hagen scheme was based on looking at the newer "grouped" Dose-symmetric script, which does have even increments on each side of zero. The previous implementation was actually correct for the original published dose symmetric script, which goes 0, +, -, -, +, ....+,-,-,+ ** This shouldn't be a big deal, so I'll leave it for now and add an option in the next version to specify grouped or original.

Bug Fixes:

Astigmatism was measured but not recorded or used.
Hagen scheme was setting up the exposure offset by one. (Zero is not included in either branch now)
Exposure filter corrected to work with non-square images
Bug since 1.1.0 in interpolation when symmetry is applied, resulting in NaNs for extrapolated values. Fixed to zero.
Set remove neighbors to save a new model when run with a project already initialized. This is only for viewing, editing it will not do anything yet. This is as opposed to running prior to initializing the project (passing the unbinned pixel size in angstroms rather than a parameter file.)
Minor changes in anticipation of switching to MCR 2018b ... str2double returns a NaN rather than an error, which buggered up a few logical checks.

Features:

Switched the WGT calculation to internally work with half transforms, and other changes to get a speed up of ~ a factor of 4. The full transform is still used prior to emClarity_2.0

2018-Dec-19

The new binary for emClarity_1.1.2 is available, and follows the format you are used to, only now you unzip it rather than pulling it down from git

2018-Nov-14

I'm into the final stages of verifying the new update is working, including:

phase plate data
new option for ctf estimation adaption of ctffind4 including an automatic handedness check
more accurate 3D-CTF (Ewald sphere) correction
fixes for the newer options for ignoring select tilts and automatic generation of exposure statistics (no need for order file.)
a number of important bug fixes, including one that was reducing the quality of the anisotropic filtering. Here's the expected result (map & one of the half-set masks) obtained for the yeast 80s tutorial data from EMPIAR 10045.

I've uploaded the source code to the repo, which you are welcome to have a look at and tell me where I should have done things better ; )

--> You may need to move or delete your mcr cache, usually found at "~/.mcrcache9.3/emClar0"

--> If anybody has the newest arch (Volta, i.e. titan V or V100) please note that matlab has not caught up yet, and unless you change some settings, will compile ALL the GPU code every time you run a program which is awful. While I'll get some notes up on this eventually, please let me know if this is an immediate issue for you.

-Ben