SPEQ: quality assessment of peptide tandem mass spectra with deep learning

This repository contains the source code for SPEQ, a tool for assessing the quality of peptide tandem mass spectra using deep learning. This work has been published in the Bioinformatics journal: https://doi.org/10.1093/bioinformatics/btab874

Dependencies

The following dependencies are needed in order to use this tool:

• TensorFlow
• Pytemoics
• NumPy
• Matplotlib
• Pandas
• Scikit-learn

Usage

To run the program, execute the SPEQ.py script using Python 3:

python3 SPEQ.py

Data format

Input data:

• The MS input data should be in MGF format.
• The data file is then automatically generated by the program.

Labels:

• A database search result in .tsv format (MS-GF+ output) is used to generate the labels.
• The decoy prefix (used in the database-searching) must be defined in the config.yml file. The program uses 1% FDR threshold.
• Alternatively, you can provide your own labels.csv file. It must have two columns:
  • Title: Corresponds to the spectrum title in the MGF file
  • Label: 0 (low quality) or 1 (high quality)