/dlomix

Python framework for Deep Learning in Proteomics

Primary LanguagePythonMIT LicenseMIT

DLOmix

Docs Build PyPI

DLOmix is a Python framework for Deep Learning in Proteomics. Initially built on top of TensorFlow/Keras, support for PyTorch can however be integrated once the main API is established.

Usage

Experiment a simple retention time prediction use-case using Google Colab    Colab

A version that includes experiment tracking with Weights and Biases is available here    Colab

Resources Repository

More learning resources can be found in the dlomix-resources repository.

Installation

Run the following to install:

$ pip install dlomix

If you would like to use Weights & Biases for experiment tracking and use the available reports for Retention Time under /notebooks, please install the optional wandb python dependency with dlomix by running:

$ pip install dlomix[wandb]

General Overview

  • data: structures for modeling the input data, processing functions, and feature extractions based on Hugging Face datasets Dataset and DatasetDict
  • eval: classes for evaluating models and reporting results
  • layers: custom layers used for building models, based on tf.keras.layers.Layer
  • losses: custom losses to be used for training with model.fit()
  • models: common model architectures for the relevant use-cases based on tf.keras.Model to allow for using the Keras training API
  • pipelines: an exemplary high-level pipeline implementation
  • reports: classes for generating reports related to the different tasks
  • constants.py: constants and configuration values

Use-cases

  • Retention Time Prediction:

    • a regression problem where the retention time of a peptide sequence is to be predicted.

  • Fragment Ion Intensity Prediction:

    • a multi-output regression problem where the intensity values for fragment ions are predicted given a peptide sequence along with some additional features.

  • Peptide Detectability (Pfly) [4]:

    • a multi-class classification problem where the detectability of a peptide is predicted given the peptide sequence.

To-Do

Functionality:

  • integrate prosit
  • integrate hugging face datasets
  • extend data representation to include modifications
  • add PTM features
  • add residual plots to reporting, possibly other regression analysis tools
  • output reporting results as PDF
  • refactor reporting module to use W&B Report API (Retention Time)
  • add additional detectability task
  • extend pipeline for different types of models and backbones
  • extend pipeline to allow for fine-tuning with custom datasets

Package structure:

  • integrate deeplc.py into models.py, preferably introduce a package structure (e.g. models.retention_time)
  • add references for implemented models in the ReadMe
  • introduce formatting and precommit hooks
  • plan documentation (sphinx and readthedocs)
  • refactor following best practices for cleaner install

Developing DLOmix

To install dlomix, along with the tools needed to develop and run tests, run the following command in your virtualenv:

$ pip install -e .[dev]

References:

[Prosit]

[1] Gessulat, S., Schmidt, T., Zolg, D. P., Samaras, P., Schnatbaum, K., Zerweck, J., ... & Wilhelm, M. (2019). Prosit: proteome-wide prediction of peptide tandem mass spectra by deep learning. Nature methods, 16(6), 509-518.

[DeepLC]

[2] DeepLC can predict retention times for peptides that carry as-yet unseen modifications Robbin Bouwmeester, Ralf Gabriels, Niels Hulstaert, Lennart Martens, Sven Degroeve bioRxiv 2020.03.28.013003; doi: 10.1101/2020.03.28.013003

[3] Bouwmeester, R., Gabriels, R., Hulstaert, N. et al. DeepLC can predict retention times for peptides that carry as-yet unseen modifications. Nat Methods 18, 1363–1369 (2021). https://doi.org/10.1038/s41592-021-01301-5

[Detectability - Pfly]

[4] Abdul-Khalek, N., Picciani, M., Wimmer, R., Overgaard, M. T., Wilhelm, M., & Gregersen Echers, S. (2024). To fly, or not to fly, that is the question: A deep learning model for peptide detectability prediction in mass spectrometry. bioRxiv, 2024-10.