incorrectly stitched denoised tomogram.mrc #29
Comments
jychoi0616
changed the title
|
For your information, I'm attaching how the denoised tomogram looks like in the flipped view. The first one is the denoised tomogram output when I used tomogram with flipped volume, and the second one is the tomogram with non flipped volume. |
|
This is confusing. Could you post the installed packages? You can get the list by running |
|
Hi, below is the list. Name Version Build Channel_libgcc_mutex 0.1 conda_forge conda-forge |
|
Hi again, I think it may have something to do with the map mode and no density information of the Aretomo-generated tomogram..? Below is the header of the tomogram generated by Aretomo. (c.f. tomogram generated by eTomo contains information for the density and map mode is 1 (16-bit).) Number of columns, rows, sections ..... 682 682 200 |
|
Hi, we also met the same problem when using the tomograms reconstructed by AreTomo. The filtered tomograms were stitched incorrectly and the 3dmod failed to open them except adding additional option "-J 1". Here is the list of corresponding building channel. packages in environment at /scopem/prog/anaconda3-5.3.1: Best, Jingwei |
|
Hi, I also can reproduce the same issue. Some extra info: using the mrc = mrcfile.open('denoised.mrc')File .../mrcfile/mrcinterpreter.py:342, in MrcInterpreter._read_data(self)
340 return
341 else:
--> 342 raise ValueError(msg)
344 self._data = np.frombuffer(data_arr, dtype=dtype).reshape(shape)
345 self._data.flags.writeable = not self._read_only
ValueError: Expected 1383888480 bytes in data block but could only read 1383838560However, rec.array((958, 926, 390, 2, 0, 0, 0, 958, 926, 390, (16443.111, 15893.863, 6693.96), (90., 90., 90.), 1, 2, 3, 0., 0., 0., 0, 49920, b'\xA0\xC0\x00\x00\x00\x00\x00\x00', b'AGAR', 20140, b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x20\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xB4\x42\x00\x00\xB4\x42\x00\x00\xB4\x42', (0., 0., 0.), b'MAP', [68, 68, 0, 0], 0., 0, [b'', b'', b'', b'', b'', b'', b'', b'', b'cryoCARE 29-Sep-22 17:28:47 ', b'']),
dtype=[('nx', '<i4'), ('ny', '<i4'), ('nz', '<i4'), ('mode', '<i4'), ('nxstart', '<i4'), ('nystart', '<i4'), ('nzstart', '<i4'), ('mx', '<i4'), ('my', '<i4'), ('mz', '<i4'), ('cella', [('x', '<f4'), ('y', '<f4'), ('z', '<f4')]), ('cellb', [('alpha', '<f4'), ('beta', '<f4'), ('gamma', '<f4')]), ('mapc', '<i4'), ('mapr', '<i4'), ('maps', '<i4'), ('dmin', '<f4'), ('dmax', '<f4'), ('dmean', '<f4'), ('ispg', '<i4'), ('nsymbt', '<i4'), ('extra1', 'V8'), ('exttyp', 'S4'), ('nversion', '<i4'), ('extra2', 'V84'), ('origin', [('x', '<f4'), ('y', '<f4'), ('z', '<f4')]), ('map', 'S4'), ('machst', 'u1', (4,)), ('rms', '<f4'), ('nlabl', '<i4'), ('label', 'S80', (10,))]) |
|
I tried to fix the headers of the halfmaps with: with mrcfile.mmap('odd.mrc', mode='r+') as mrc:
mrc.update_header_from_data()
mrc.update_header_stats()But it did't help. |
|
Hi Tim-Oliver and Thorsten, I've made the cryoCARE workflow inside Scipion using two different algorithms for the even/odd tomograms reconstruction: aretomo and imod. With the first one, I experienced the problems reported here (see details below), but I had no problem regarding the predicted denoised tomogram if working with imod. I hope it helps. This is what I get when I try to load it with Imod 3dmod: And with xmipp: It seems that a very small portion of the data is lost, but I don't know if there's a relation with the aretomo reconstruction. Best, Jorge |
|
Can you share such a corrupted halfmap? Best, |
|
Hi @thorstenwagner, These are the links for:
I hope that helps! |
|
Now with the Aretomo 1.3 version, this bug was fixed. So now the density information doesn't get lost. Could that be cryoCARE was designed to be used with tomograms of 16 bits rather than 32 bits? |
|
Hi, The problem that I noticed here arises from copying parts of the mrc header that are not the same as the newly saved data. In the aretomo generated tomograms, there are some extra bytes in the mrc files, but in cryoCARE output, those bytes are not written into the file, we only have the necessary data. Still, the header indicates those bytes are there and should be skipped when we open the file which means we cannot read the file properly since then we miss data. I managed to solve the problem by skipping to save the information of extra bytes in the header. |
|
Hi, thank you!
|
|
Hi @jychoi0616, this is the modified denoise function that worked (from the file https://github.com/juglab/cryoCARE_pip/blob/master/cryocare/scripts/cryoCARE_predict.py): |
Thanks a lot! This worked :) |
Dear the team,
I get errors to open denoised tomogram.mrc file (as the attached) using 3dmod. I managed to open the denoised tomogram using a command 3dmod -J 1, but the tomogram seemed it has been stitched together incorrectly. And also it seems this issue was specific to tomograms generated by AreTomo. I guess this issue could be connected the issue #26.
Best wishes,
Joy
The text was updated successfully, but these errors were encountered: