AlphaPeptDeep (PeptDeep)

The peptdeep.pipeline_api module provides high-level entry points for common deep learning workflows in proteomics. It currently supports:

• Transfer Learning: Use transfer_learn to fine-tune models.
• Library Prediction: Use generate_library to predict properties for a peptide library.
• Rescoring: Use rescore to improve peptide identifications using deep learning models.

Note that the default parameters for these functions are controlled globally via peptdeep.settings.global_settings.

The peptdeep.spec_lib.library_factory module provides library classes designed to predict spectral libraries. These classes can generate predictions based on various input types, including:

• Peptides (sequences)
• Modified peptides
• Precursors
• Proteins (via FASTA files)

For users who want a standalone tool without Python or CLI knowledge, you can download pre-compiled installers from the GitHub releases page.

Platform Specifics:

• Windows: Download the .exe (e.g., peptdeep-X.Y.Z-windows-amd64.exe) and run it. Always install into a new folder, as the installer does not overwrite existing installations properly.
• Linux: Download the .deb file and install via dpkg -i.
• MacOS: Download the .pkg matching your architecture (arm64 for M1/M2/M3, x64 for Intel).
  • Note: On newer macOS versions, you may need to allow the app in System Settings > Privacy & Security > Security by clicking Open Anyway if it is blocked.

To use peptdeep as a Python package in an existing environment (e.g., a Jupyter notebook), use pip. Recommended Python versions are 3.8 or 3.9.

Standard installation: Installs the core package. This avoids strict versioning conflicts with other tools but does not guarantee dependency compatibility if new versions are released.

Stable installation: Forces peptdeep to use dependency versions known to be compatible.

GUI installation: Installs the Graphical User Interface version via pip.

Note on Raw Data: To access Thermo or Sciex raw data, you must also install AlphaRaw (which depends on Mono for Mac/Linux).

# Standard installation
pip install peptdeep

# Stable installation
pip install "peptdeep[stable]"

# Install GUI version
pip install "peptdeep[gui]"

# Install stable GUI version
pip install "peptdeep[stable,gui-stable]"

If you are contributing to the repository, it is highly recommended to use the provided pre-commit hooks. The CI pipeline enforces these checks before any branch can be merged.

To install the hooks once:

pre-commit install

To manually run all checks on all files:

pre-commit run --all-files

To enable GPU acceleration, use the Pip or Developer installation methods and manually upgrade to a GPU-enabled version of PyTorch. Ensure your NVIDIA drivers are compatible by checking nvidia-smi.

pip install torch --extra-index-url https://download.pytorch.org/whl/cu116 --upgrade

Use peptdeep.spec_lib.translate to convert alphabase spectral libraries into TSV formats compatible with tools like DiaNN, OpenSWATH, and Spectronaut.

Note: This module is considered legacy and will be replaced by alphabase.spectral_library.translate. For the most up-to-date implementation, refer to the alphabase repository.

For a detailed walkthrough, see the tutorial: Tutorial: translate spectral libraries <../nbs/alphapeptdeep_hdf_to_tsv>.

To modify the source code or access all features, install in "editable" mode using a Conda environment. This is recommended for high-throughput experiments as the developer version often outperforms precompiled versions.

1. Clone the repository.
2. Create and activate a Conda environment (Python 3.9 recommended).
3. Install with the [development] extra to include development dependencies.

# Clone the repo
git clone https://github.com/MannLabs/alphapeptdeep.git && cd alphapeptdeep

# Setup Conda environment
conda create --name peptdeep python=3.9 -y
conda activate peptdeep

# Install in editable mode with development dependencies
pip install -e ".[development]"

The peptdeep.protein.fasta module provides the primary API for generating spectral libraries directly from FASTA files (peptide lists). This allows you to predict properties for a set of peptides derived from a protein sequence database.

For a complete walkthrough of the workflow, refer to the Tutorial: spectral library from fasta notebook.