DeepGlyco is a framework for predicting fragment mass spectra of intact glycopeptides.
For model training, NVIDIA graphics cards with CUDA are recommended.
The following software and packages are required:
- Python (version 3.9.16, Anaconda distribution is recommended)
- PyTorch (version 1.12.1)
- DGL (version 1.0.1)
- numpy (version 1.23.5)
- pandas (version 1.5.2)
- scipy (version 1.10.1)
- scikit-learn (version 1.2.1)
- statsmodels (version 0.13.2)
- h5py (version 3.7.0)
- pymzml (version 2.5.2)
The following packages are optional for visualization:
- matplotlib (version 3.6.2)
- networkx (version 3.0)
- tensorboard (version 2.10.0)
Later versions may be compatible, but have not been tested.
Tutorials are avaliable in the docs
folder.
DeepGlyco Tutorial: Training New Models for MS/MS describes the analysis workflow for training MS/MS models.
DeepGlyco Tutorial: Model Finetuning with New Data describes the analysis workflow for finetuning MS/MS models using users' data.
DeepGlyco Tutorial: Training New Models for iRT describes the analysis workflow for training iRT models.
DeepGlyco Tutorial: DIA Analysis Using Predicted Libraries describes the DIA analysis workflow using GproDIA and spectral libraries predicted by DeepGlyco.
DeepGlyco Tutorial: Rescoring Glycopeptide Structures describes the analysis workflow for rescoring glycopeptide spectral matches reported by other tools using spectral libraries predicted by DeepGlyco.
DeepGlyco is distributed under a BSD license. See the LICENSE file for details.