alphatims 0.1.210323

A Python package for Bruker TimsTOF raw data accession and visualization

---

AlphaTims

AlphaTims is an open-source Python package that provides fast accession and visualization of unprocessed LC-TIMS-Q-TOF data from Bruker’s timsTOF Pro instruments. It indexes the data such that it can easily be sliced along all five dimensions: LC, TIMS, QUADRUPOLE, TOF and DETECTOR. It was developed by the Mann Labs at the Max Planck Institute of Biochemistry. To enable all hyperlinks in this document, please view it at https://github.com/MannLabs/alphatims.

• AlphaTims
  • About
  • License
  • Installation
    • One-click GUI
    • Pip installer
    • Developer installer
    • Installation issues
  • Test data
    • Test sample
    • LC
    • DDA
    • DIA
  • Usage
    • GUI
    • CLI
    • Python and jupyter notebooks
    • Other tools
  • Performance
    • Speed
    • RAM
  • Troubleshooting
  • How it works
    • Bruker raw data
    • TimsTOF objects in Python
    • Slicing TimsTOF objects
  • Future perspectives
  • How to contribute

---

About

High-resolution quadrupole time-of-flight (Q-TOF) tandem mass spectrometry can be coupled to several other analytical techniques such as liquid chromatography (LC) and trapped ion mobility spectrometry (TIMS). LC-TIMS-Q-TOF has gained considerable interest since the introduction of the Parallel Accumulation–Serial Fragmentation (PASEF) method in both data-dependent (DDA) and data-independent acquisition (DIA). With this setup, ion intensity values are acquired as a function of the chromatographic retention time, ion mobility, quadrupole mass to charge and TOF mass to charge. As these five-dimensional data points are detected at GHz rates, datasets often contain billions of data points which makes them impractical and slow to access. Raw data are therefore frequently binned for faster data analysis or visualization. In contrast, AlphaTims is a Python package that provides fast accession and visualization of unprocessed raw data. By recognizing that all measurements are ultimately arrival times linked to intensity values, it constructs an efficient set of indices such that raw data can be interpreted as a sparse five-dimensional matrix. On a modern laptop, this indexing takes less than half a minute for raw datasets of more than two billion datapoints. Following this step, interactive visualization of the same dataset can also be done in milliseconds. AlphaTims is freely available, open-source and available on all major Operating Systems. It can be used with a graphical user interface (GUI), a command-line interface (CLI) or as a regular Python package.

---

License

AlphaTims was developed by the Mann Labs at the Max Planck Institute of Biochemistry and is freely available with an Apache License. Since AlphaTims uses Bruker libraries (available in the alphatims/ext folder) and external Python packages (available in the requirements folder), additional third-party licenses are applicable.

---

Installation

AlphaTims can be installed and used on all major operating systems (Windows, MacOS and Linux). There are three different types of installation possible:

• One-click GUI installer: Choose this installation if you only want the GUI and/or keep things as simple as possible.
• Pip installer: Choose this installation if you want to use AlphaTims as a Python package in an existing Python 3.8 environment (e.g. a Jupyter notebook). If needed, the GUI and CLI can be installed with pip as well.
• Developer installer: Choose this installation if you are familiar with CLI tools, conda and Python. This installation allows access to all available features of AlphaTims and even allows to modify its source code directly. Generally, the developer version of AlphaTims outperforms the precompiled versions which makes this the installation of choice for high-throughput experiments.

IMPORTANT: While AlphaTims is mostly platform independent, some calibration functions require Bruker libraries which are only available on Windows and Linux.

One-click GUI

The GUI of AlphaTims is a completely stand-alone tool that requires no knowledge of Python or CLI tools. Click on one of the links below to download the latest release for:

• Windows
• MacOS
• Linux

IMPORTANT: Please refer to the GUI manual for detailed instructions on the installation, troubleshooting and usage of the stand-alone AlphaTims GUI.

Older releases remain available on the release page, but no backwards compatibility is guaranteed.

Pip

AlphaTims can be installed in an existing Python 3.8 environment with a single bash command. This bash command can also be run directly from within a Jupyter notebook by prepending it with a !. The lightweight version of AlphaTims that purely focuses on data accession (no plotting without additional packages) can be installed with:

pip install alphatims

Alternatively, some basic plotting functions and the complete GUI can be installed with the command:

pip install alphatims[plotting]

This assumes git is accessible to this environment. If this is not the case, it can often be installed in the environment with the command conda install git -y. When a new version of AlphaTims becomes available, the old version can easily be upgraded by running the command again with an additional --upgrade flag:

pip install alphatims --upgrade

Developer

AlphaTims can also be installed in developer mode with a few bash commands. This allows to fully customize the software and modify the source code to your specific needs. For any Python package, it is highly recommended to use a conda virtual environment. Once conda is downloaded and installed, an AlphaTims environment can be created and activated with:

conda create -n alphatims python=3.8 pip=20.2 -y
conda activate alphatims

When an editable Python package is installed, its source code is stored in a transparent location of your choice. While optional, it is advised to first create a specific folder for AlphaTims and navigate to this folder:

mkdir ~/folder/where/to/install/software
cd ~/folder/where/to/install/software

Next, download the AlphaTims repository from GitHub either directly or with the git command (installable with conda install git -y):

git clone https://github.com/MannLabs/alphatims.git

This creates a new AlphaTims subfolder in the current directory. The following commands assume you did not perform an additional cd alphatims to change the current directory to this subfolder. Next, AlphaTims and all dependancies need to be installed. To take advantage of all features and allow development, this is best done by installing both the plotting dependencies and development dependencies instead of only the core dependencies:

pip install -e './alphatims[plotting,development]'

By using the editable flag -e, all modifications to the AlphaTims source code folder are directly reflected when running AlphaTims. Note that the AlphaTims folder cannot be moved and/or renamed if an editable version is installed.

The following steps are optional, but make working with AlphaTims slightly more convenient:

• To avoid calling conda activate alphatims and conda deactivate every time AlphaTims is used (this is optional), the binary execution (which still reflects all modifications to the source code) can be added as an alias. On linux and MacOS, this can be done with e.g.:

  conda activate alphatims
  alphatims_bin="$(which alphatims)"
  echo "alias alphatims='"${alphatims_bin}"'" >> ~/.bashrc
  conda deactivate
  

  When zsh is the default terminal instead of bash, replace ~/.bashrc with ~/.zshrc. On Windows, the command where alphatims can be used to find the location of the binary executable. This path can then be (permanently) added to Windows' path variable.
• When using Jupyter notebooks and multiple conda environments direcly from the terminal, it is recommended to conda install nb_conda_kernels in the conda base environment. Hereafter, running a jupyter notebook from the conda base environment should have a python [conda env: alphatims] kernel available, in addition to all other conda kernels in which the command conda install ipykernel was run.

Installation issues

See the general troubleshooting section.

---

Test data

AlphaTims is compatible with both ddaPASEF and diaPASEF.

Test sample

A test sample of human cervical cancer cells (HeLa, S3, ATCC) is provided for AlphaTims. These cells were cultured in Dulbecco's modified Eagle's medium (all Life Technologies Ltd., UK). Subsequently, the cells were collected, washed, flash-frozen, and stored at -80 °C. Following the previously published in-StageTip protocol, cell lysis, reduction, and alkylation with chloroacetamide were carried out simultaneously in a lysis buffer (PreOmics, Germany). The resultant dried peptides were reconstituted in water comprising 2 vol% acetonitrile and 0.1% vol% trifluoroacetic acid, yielding a 200 ng/µL solution. This solution was further diluted with water containing 0.1% vol% formic acid. The manufacturer's instructions were followed to load approximately 200ng peptides onto Evotips (Evosep, Denmark).

LC

Single-run LC-MS analysis was executed via an Evosep One LC system (Evosep). This was coupled online with a hybrid TIMS quadrupole TOF mass spectrometer (Bruker timsTOF Pro, Germany). A silica emitter (Bruker) was placed inside a nano-electrospray ion source (Captive spray source, Bruker) and connected to an 8 cm x 150 µm reverse phase column to perform LC. The column was packed with 1.5 µm C18-beads (Pepsep, Denmark). Mobile phases were water and acetonitrile, buffered with 0.1% formic acid. The samples were separated with a predefined 60 samples per day method (Evosep).

DDA

A ddaPASEF dataset is available for download from the release page. Each topN acquisition cycle consisted of 10 PASEF MS/MS scans, and the accumulation and ramp times were set to 100 ms. Single-charged precursors were excluded using a polygon filter in the m/z-ion mobility plane. Furthermore, all precursors, which reached the target value of 20000, were excluded for 0.4 min from the acquisition. Precursors were isolated with a quadrupole window of 2 Th for m/z <700 and 3 Th for m/z >700.

DIA

The same sample was also acquired with diaPASEF and is also available for download from the release page. The "high-speed" method (mass range: m/z 400 to 1000, 1/K0: 0.6 – 1.6 Vs cm- 2, diaPASEF windows: 8 x 25 Th) was used, as described in Meier et al.

---

Usage

There are three ways to use AlphaTims:

• GUI: This allows to browse and visualize the data.
• CLI: This allows to incorporate AlphaTims in automated workflows.
• Python: This allows to access data and explore it interactively.

The first time you use a fresh installation of AlphaTims is often quite slow because some functions might still need compilation on you local operating system and architecture. Subsequent use should be a lot faster.

GUI

Please refer to the GUI manual for detailed instructions on the installation, troubleshooting and usage of the stand-alone AlphaTims GUI.

If the GUI was not installed through a one-click GUI installer, it can be activate with the following bash command:

alphatims gui

Note that this needs to be prepended with a ! when you want to run this from within a Jupyter notebook. When the command is run directly from the command-line, make sure you use the right environment (activate it with e.g. conda activate alphatims or set an alias to the binary executable).

CLI

The CLI can be run with the following command (after activating the conda environment with conda activate alphatims or if an alias was set to the alphatims executable):

alphatims -h

It is possible to get help about each function and their (required) parameters by using the -h flag. For instance, the command alphatims export hdf -h will produce the following output:

************************
* AlphaTims 0.0.210310 *
************************
Usage: alphatims export hdf [OPTIONS] BRUKER_D_FOLDER

  Export BRUKER_D_FOLDER as hdf file.

Options:
  --disable_overwrite            Disable overwriting of existing files.
  --enable_compression           Enable compression of hdf files. If set, this
                                 roughly halves files sizes (on-disk), at the
                                 cost of taking 2-10 longer accession times.
  -o, --output_folder DIRECTORY  A directory for all output (blank means
                                 `input_file` root is used).

  -l, --log_file PATH            Save all log data to a file (blank means
                                 'log_[date].txt' with date format
                                 yymmddhhmmss in 'log' folder of AlphaTims
                                 directory).  [default: ]
  -t, --threads INTEGER          The number of threads to use (0 means all,
                                 negative means how many threads to leave
                                 available).  [default: -1]
  -s, --disable_log_stream       Disable streaming of log data.
  -p, --parameter_file FILE      A .json file with (non-required) parameters
                                 (blank means default parameters are used).
                                 NOTE: Parameters defined herein override all
                                 default and given CLI parameters.
  -e, --export_parameters FILE   Save currently selected parameters to a
                                 parameter file.
  -h, --help                     Show this message and exit.

For this particular command, the line Usage: alphatims export hdf [OPTIONS] BRUKER_D_FOLDER shows that you always need to provide a path to a BRUKER_D_FOLDER and that all other options are optional (indicated by the brackets in [OPTIONS]). Each option can be called with a double dash -- followed by a long name, while common options also can be called with a single dash - followed by their short name. It is indicated what type of parameter is expected, e.g. a DIRECTORY for --output_folder or nothing for enable/disable flags. Defaults are also shown and all parameters will be saved in a log file. Alternatively, all used parameters can be exported with the --export_parameters option and the non-required ones can be reused with the --parameter_file.

Python and Jupyter notebooks

AlphaTims can be imported as a Python package into any Python script or notebook with the command import alphatims. Documentation for all functions is available in the Read the Docs API.

A brief Jupyter notebook tutorial on how to use the API is also present in the nbs folder. When running locally it provides interactive plots, which are not rendered on GitHub. Instead, they are available as individual html pages in the nbs folder.

Other tools

• Initial exploration of Bruker TimsTOF data files can be done by opening the .tdf file in the .d folder with an SQL browser.
• HDF files can be explored with HDF Compass or HDFView.

---

Performance

Performance can be measured in function of speed or RAM usage.

Speed

Typical time performance statistics on data in-/output and slicing of standard HeLa datasets include:

type	gradient	datapoints	reading (raw/HDF)	export HDF	slicing (in ms)
DDA	6 min	214,172,697	1.48 s / 445 ms	569 ms	1.85 / 48 / 27.1 / 89.7
DIA	6 min	158,552,099	1.06 s / 295 ms	409 ms	7.35 / 24.1 / 649 / 110
DDA	21 min	295,251,252	3.07 s / 755 ms	779 ms	2.17 / 68.3 / 106 / 181
DIA	21 min	730,564,765	5.24 s / 1.90 s	1.79 s	0.860 / 123 / 4910 / 392
DDA	120 min	2,074,019,899	23.2 s / 10.1 s	4.94 s	0.671 / 338 / 557 / 1100

All slices were performed in a single dimension. Including more slices makes the analysis more stringent and hence faster. The considered dimensions were:

• LC: 100.0 <= retention_time < 100.5
• TIMS: scan_index = 450
• Quadrupole: 700.0 <= quad_mz_values < 710.0
• TOF: 621.9 <= tof_mz_values < 622.1

All of these analyses were timed with timeit and are the average of at least 7 runs. They were obtained on the following system:

• MacBook Pro: 13-inch, 2020, Four Thunderbolt 3 ports
• OS version: macOS Catalina 10.15.7
• Processor: 2.3 GHz Quad-Core Intel Core i7
• Memory: 32 GB 3733 MHz LPDDR4X
• Startup Disk: Macintosh HD

Full details are available in the performance notebook.

RAM

On average, RAM usage is twice the size of a raw Bruker .d folder. Since most .d folders have file sizes of less than 10 Gb, a modern computer with 32 Gb RAM suffices to explore most datasets with ease.

---

Troubleshooting

Common installation/usage issues include:

• Always make sure you have activated the AlphaTims environment with conda activate alphatims. If this fails, make sure you have installed conda and have created an AlphaTims environment with conda create -n alphatims python=3.8.
• No git command. Make sure git is installed. In a notebook !conda install git -y might work.
• Wrong Python version. AlphaTims is only guaranteed to be compatible with Python 3.8. You can check if you have the right version with the command python --version (or !python --version in a notebook). If not, reinstall the AlphaTims environment with conda create -n alphatims python=3.8.
• Dependancy conflicts/issues. Pip changed their dependancy resolver with pip version 20.3. Downgrading pip to version 20.2 with pip install pip==20.2 (before running pip install alphatims) could solve dependancy conflicts.
• AlphaTims is not found. Make sure you use the right folder. Local folders are best called by prefixing them with ./ (e.g. pip install ./alphatims). On some systems, installation specifically requires (not) to use single quotes ' around the AlphaTims folder, e.g. pip install './alphatims[plotting,development]'.
• Modifications to the AlphaTims source code are not reflected. Make sure you use the -e flag when using pip install -e alphatims.
• Numpy does not work properly. On Windows, numpy==1.19.4 has some issues. After installing AlphaTims, downgrade NumPy with pip install numpy==1.19.3.
• Exporting PNG images with the CLI or Python package might not work out-of-the-box. If a conda environment is used, this can be fixed by running conda install -c conda-forge firefox geckodriver in the AlphaTims conda environment. Alternatively, a file can be exported as html and opened in a browser. From the browser there is a save as png button available.

---

How it works

The basic workflow of AlphaTims looks as follows:

• Read data from a Bruker .d folder.
• Convert data to a TimsTOF object in Python and store them as a persistent HDF5 file.
• Use Python's slicing mechanism to retrieve data from this object e.g. for visualisation.

Bruker raw data

Bruker stores TimsTOF raw data in a .d folder. The two main files in this folder are analysis.tdf and analysis.tdf_bin.

The analysis.tdf file is an SQL database, in which all metadata are stored together with summarised information. This includes the Frames table, wherein information about each individual TIMS cycle is summarised including the retention time, the number of scans (i.e. a single TOF push is related to a single ion mobility value), the summed intensity and the total number of ions that have hit the detector. More details about individual scans of the frames are available in the PasefFrameMSMSInfo (for PASEF acquisition) or DiaFrameMsMsWindows (for diaPASEF acquisition) tables. This includes quadrupole and collision settings of the frame/scan combinations.

The analysis.tdf_bin file is a binary file that contains the number of detected ions per individual scan, all detector arrival times and their intensity values. These values are grouped and compressed per frame (i.e. TIMS cycle), thereby allowing fast appendage during online acquisition.

TimsTOF objects in Python

AlphaTims first reads relevant metadata from the analysis.tdf SQL database and creates a Python object of the bruker.TimsTOF class. Next, AlphaTims reads the summary information from the Frames table and creates three empty arrays:

• An empty tof_indices array, in which all TOF arrival times of each individual detector hit will be stored. Its size is determined by summing the number of detector hits for all frames.
• An empty intensities array of the same size, in which all intensity values of each individual detector hit will be stored.
• An empty tof_indptr array, that will store the number of detector hits per scan. Its size is equal to (frame_max_index + 1) * scans_max_index + 1. It includes one additional frame to compensate for the fact that Bruker arrays are 1-indexed, while Python uses 0-indexing. The final +1 is because this array will be converted to an offset array, similar to the index pointer array of a compressed sparse row matrix. Typical values are scans_max_index = 1000 and frame_max_index = gradient_length_in_seconds * 10, resulting in approximately len(tof_indptr) = 10000 * gradient_length_in_seconds.

After reading the PasefFrameMSMSInfo or DiaFrameMsMsWindows table from the analysis.tdf SQL database, four arrays are created:

• A quad_indptr array that indexes the tof_indptr array. Each element points to an index of the tof_indptr where the voltage on the quadrupole and collision cell is adjusted. For PASEF acquisitions, this is typically 20 times per MSMS frame (turning on and off a value for 10 precursor selections) and once per change from an MS (precursor) frame to an MSMS (fragment) frame. For diaPASEF, this is typically twice to 10 times per frame and with a repetitive pattern over the frame cycle. This results in an array of approximately len(quad_indptr) = 100 * gradient_length_in_seconds. As with the tof_indptr array, this array is converted to an offset array with size +1.
• A quad_low_values array of len(quad_indptr) - 1. This array stores the lower m/z boundary that is selected with the quadrupole. For precursors without quadrupole selection, this value is set to -1.
• A quad_high_values array, similar to quad_low_values.
• A precursor_indices array of len(quad_indptr) - 1. For PASEF this array stores the index of the selected precursor. For diaPASEF, this array stores the WindowGroup of the fragment frame. A value of 0 indicates an MS1 ion (i.e. precursor) without quadrupole selection.

After processing this summarising information from the analysis.tdf SQL database, the actual raw data from the analysis.tdf_bin binary file is read and stored in the empty tof_indices, intensities and tof_indptr arrays.

Finally, three arrays are defined that allow quick translation of frame_, scan_ and tof_indices to rt_values, mobility_values and mz_values arrays.

• The rt_values array is read read directly from the Frames table in analysis.tdf and has a length equal to frame_max_index + 1. Note that an empty zeroth frame with rt = 0 is created to make Python's 0-indexing compatible with Bruker's 1-indexing.
• The mobility_values array is defined by using the function tims_scannum_to_oneoverk0 from timsdata.dll on the first frame and typically has a length of 1000.
• Similarly, the mz_values array is defined by using the function tims_index_to_mz from timsdata.dll on the first frame. Typically this has a length of 400000.

All these arrays can be loaded into memory, taking up roughly twice as much RAM as the .d folder on disk. This increase in RAM memory is mainly due to the compression used in the analysis.tdf_bin file. The HDF5 file can also be compressed so that its size is roughly halved and thereby has the same size as the Bruker .d folder, but (de)compression reduces accession times by 3-6 fold.

Slicing TimsTOF objects

Once a Python TimsTOF object is available, it can be loaded into memory for ultrafast accession. Accession of the data object is done by simple Python slicing such as e.g. selected_ion_indices = data[frame_selection, scan_selection, quad_selection, tof_selection]. This slicing returns a pd.DataFrame for subsequent analysis. The columns of this dataframe contain all information for all selected ions, i.e. frame, scan, precursor and tof indices and rt, mobility, quad_low, quad_high, mz and intensity values. See the tutorial jupyter notebook for usage examples.

---

Future perspectives

• Detection of:
  • precursor and fragment ions
  • isotopic envelopes (i.e. features)
  • fragment clusters (i.e. pseudo MSMS spectra)

---

How to contribute

All contributions are welcome. Feel free to post a new issue or clone the repository and create a PR with a new branch. For more information see the Contributors License Agreement