Version: 1.1.3 build -228
DIAtensor is software tool to convert mzXML of data-independent acquisition (DIA) mass spectrometry (MS) to DIA tensor. This tool is based on Python, which can read mzXML mass spectrometry data, check the integrity of the data, pool the mass-to-charge ratio and retention time, and perform frequency division bin coloring such as intensity values. With DIA Tensor, a classification deep learning model without the need for a large amount of extraction library chromatographic peaks to identify peptides.
Software: https://github.com/guomics-lab/DIAtensor/releases
Manual: https://github.com/guomics-lab/DIAtensor/blob/master/README.md
Please also check the commands listed below.
Phenotype Classification using Proteome Data in a Data-Independent Acquisition Tensor Format
https://pubs.acs.org/doi/10.1021/jasms.0c00254?fig=tgr1&ref=pdf
Operating System:
- Windows:
- Windows 10 64-bit
- Linux:
- Centos 7.4 64-bit
- Ubuntu 18.04 64-bit
RAM:
- Minimum: 16GB
- Recommend: 32GB and upper
- Download DIAtensor.exe from github
- Open the Windows console CMD and switch to the directory including DIAtensor.exe
- Run the DIAtensor.exe with parameters(
DIAtensor.exe [paraneters])
- Download DIAtensor.tar.gz from github
- Unzip the file (
tar zxvf DIAtensor.tar.gz) - Open the Linux terminal and switch to the directory including DIAtensor
- Run the DIAtensor with parameters(
./DIAtensor [paraneters])
Raw data files: .mzXML
mzXML file convertion: the Peak Picking operation should be checked in the Msconvert software with MS Levels parameter should be set to 1-2.
- tensors (*.diat)
dtype : uint16 / uint32
dimension z : window index (from low to high)
dimension y : mz index (from low to high)
dimension x : cycle index (from low to high)
dimension color : intensity value
- images (*.png / *.bmp)
dimension y : mz index (from low to high)
dimension x : cycle index (from low to high)
- sampleId
[sampleId]
Sample name: D20181207yix_HCC_SW_T_41A-HCC_yix_9
- window
_win_[number](_[number])
MS2 window number 20: _win_20
MS2 window number 1~66: win_1_66
- gradient time
_grad_[start_minute]_[end_minute]
Effective gradient from 4 to 48 minutes: _grad_4_48
- cycle
_cycle_[number]
cycle 1400: _cycle_1400
- m/z range
_mz_[number]_[number]
_mz_400_1500
- m/z bin & pool
_bin_[number]
Binning m / z with 0.5Da accuracy: _bin_0.5
Binning m / z with 0.01 Da and main and auxiliary peak methods: _bin_peaks
- data type
_[dtype]
_uint16 or _uint32
- RT pool
_[pool method]_[pool width]_[pool height]
4 x 1 regional maximum pooling: _max_rt_4_1
- color scheme
_[color](_[color])_[number]
Black to white 255 gradient: _Gray_255
Green to red 255 gradient: _Gr_Re_255
DIAtensor.exe [commands]
Commands can be supplied in arbitrary order.
-i / --input <path>
Input address of the mzXML raw file.
-o / --output <path>
Output address of the DIAtensor result file.
-T / --out_type <’image’/’tensor’>
Output type of the DIAtensor result file: image or tensor.
-t / --threads <int>
Number of threads executing at the same time. default is 1.
-m / --mz <int> <int>
MS2 m / z range. The default range is 400~1500.
-g / --grad <float> <float>
Effective gradient time of the dataset.
-a / --auto_cycle
When this parameter is used, the cycle number alignment value is automatically calculated and aligned on the data set. This parameter will save the input dataset cycle distribution (cycle.csv / cycle_grad.csv) into the output folder. This parameter is mutually exclusive with manual_cycle.
-c / --manual_cycle <int>
This parameter records and aligns the manually selected data set cycle number. This parameter is mutually exclusive with auto_cycle.
-b / --bin <float>
This parameter records the binning accuracy of m / z (unit is Da), the default is 0.5.
--pool_mz
When this parameter is used and the binning accuracy parameter bin is 0.01 and the minimum ms2 m / z is 400, pooling of the main and auxiliary peaks in the m / z dimension will be performed.
Commands below only available if output type is ‘image’:
--pool_rt
When this parameter is used, max pooling on the RT (cycle) dimension (4x1) will be performed.
-D / --dimension <'3D'/'2D'>
This parameter determines the output image YN sub-window (3D means that the sub-window is out of the picture, 2D means that the windows are arranged in a big picture in order), the default is the sub-window(3D).
-C / --color <'view'/'gray'>
This parameter determines whether the output image is a visual image (view) or a grayscale image (gray). The default is a visual image (view).
-F / --format <'png'/'bmp'>
This parameter determines whether the output image is a png or a bmp.The default is png.
- Generation of a tensor by 1400 cycle with pooling in mz dimensions.
DIAtensor.exe -i E: /HCCSW -o E: /tensor -T tensor -c 1400 -b 0.01 --pool_mz
- Generation of a tensor by the m/z range 400~2000 Da, auto aligned cycle based on the gradient of 0-45 minutes,with pooling in mz dimensions.
DIAtensor.exe -i E: /HCCSW -o E: /tensor -T tensor -m 400 2000 -a -g 0 45 -b 0.01 --pool_mz
Replace <file_path> with your own file path, and run following codes in Python 3.6+ environment with Numpy 1.16.2+ installed. The RAM should be no less than 16GB.
import numpy as np
diat = np.load("<file_path>.diat")['diat']
- Generation of an aligned image by 1400 cycle with pooling in both the m/z and RT dimensions.
DIAtensor.exe -i E: /HCCSW -o E: /img -T image -c 1400 -b 0.01 --pool_mz --pool_rt -D 2D -C view -F png
- Generation of a set of aligned images by 1000 cycles and m/z and RT dimensions without pooling, binning with 0.5 Da accuracy.
DIAtensor.exe -i E: /HCCSW -o E: /img -T image -c 1000 -b 0.5 -D 3D -C view -F png
- Generation of a set of aligned images by the m/z range 400~2000 Da, auto aligned cycle based on the gradient of 4-48 minutes, and perform the main and auxiliary peak pooling in the m/z dimension. The RT dimension does not do pooling, and the windowed TPD dataset grayscale bmp image.
DIAtensor.exe -i E: /TPD_4_48 -o E: /img -T image -m 400 2000 -a -g 4 48 -b 0.01 --pool_mz -D 3D -C gray -F bmp
- The input and output addresses E: / HCCSW and E: / img of the sample data set need to be modified to the folder addresses in the user's computer.
- When the DIAtensor software is running, the single-process operation requires no less than 16GB of memory. It is recommended that 16GB memory computers use the default 1 thread, 32GB memory computers set threads to 2, and other memory sizes and so on.
- Where there is any difference between the software instruction and the description provided by the software help parameter (-h), one should always follow the help description provided by software.