Oktoberfest: Open-source spectral library generation and rescoring pipeline based on Prosit

Area of Science:

• Proteomics
• Computational Biology
• Bioinformatics

Background:

• Machine learning (ML) and deep learning (DL) models like Prosit enable high-quality in silico peptide property prediction.
• These predictions are crucial for applications such as data-independent acquisition (DIA) analysis and search engine result rescoring in proteomics.

Purpose of the Study:

• To introduce Oktoberfest, an open-source Python package for spectral library generation and peptide property prediction.
• To provide a search engine-agnostic tool that integrates state-of-the-art ML/DL models into proteomics analysis pipelines.
• To facilitate the adoption and local installation of advanced ML/DL tools for proteomics research.

Main Methods:

• Development of Oktoberfest, a Python package integrating spectral library generation and rescoring functionalities.
• Leveraging online peptide property predictions from ML/DL models.
• Demonstrating the package's performance on two distinct rescoring use cases.

Main Results:

• Oktoberfest successfully reproduces and enhances results from previous rescoring analyses.
• The package is search engine agnostic, promoting wider applicability.
• Successful local installation via PyPI is confirmed.

Conclusions:

• Oktoberfest democratizes access to advanced ML/DL-based peptide property prediction for proteomics.
• The open-source package streamlines spectral library generation and data analysis.
• It empowers researchers to improve the accuracy and efficiency of proteomics studies.