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

Abstract

Machine learning (ML) and deep learning (DL) models for peptide property prediction such as Prosit have enabled the creation of high quality in silico reference libraries. These libraries are used in various applications, ranging from data‐independent acquisition (DIA) data analysis to data‐driven rescoring of search engine results. Here, we present Oktoberfest, an open source Python package of our spectral library generation and rescoring pipeline originally only available online via ProteomicsDB. Oktoberfest is largely search engine agnostic and provides access to online peptide property predictions, promoting the adoption of state‐of‐the‐art ML/DL models in proteomics analysis pipelines. We demonstrate its ability to reproduce and even improve our results from previously published rescoring analyses on two distinct use cases. Oktoberfest is freely available on GitHub (https://github.com/wilhelm‐lab/oktoberfest) and can easily be installed locally through the cross‐platform PyPI Python package.

From data to discovery: The essential role of computational tools in proteomics

“…Further, Joyce and Searle's review on computational approaches for phosphoproteomics identification and localization presents the future potential of using predicted peptide properties for interpreting phosphopeptide positional isomers and disambiguating chimeric spectra containing multiple isomeric peptides that differ only in the phosphorylation location [3]. Additionally, Picciani et al introduce the Oktoberfest tool, leveraging the Prosit peptide property prediction model to create simulated spectral libraries and rescore peptidespectrum matches, thereby providing a convenient tool to use these predictions for extracting more information from mass spectrometry experiments [4]. This issue also highlights manuscripts that explore the varied applications of computational approaches in mass spectrometry.…”

moPepGen: Rapid and Comprehensive Identification of Non-canonical Peptides

“…Fragment ion matching between the experimental and theoretical spectra was performed using a similar approach to IPSA 66 . Theoretical spectra with predicted fragment peak intensities were generated using Prosit through the Oktoberfest 67 Python package using parameters (e.g., fragmentation method and energy) in accordance with the original publication 42 . Similarities between the experimental spectra and the Prosit predicted spectra were estimated using cross-correlation 68 , using the same parameters (e.g., fragment_bin_offset) during database search with Comet.…”

Machine learning‐based peptide‐spectrum match rescoring opens up the immunopeptidome

“…Existing rescoring tools mainly differ from each other by their use of distinct feature sets and prediction models. Some PSM rescoring tools use only a few features by default [30–32], while others use dozens [33, 34] to 100 features [35, 36]. Additionally, some tools allow adjusting the number of features used.…”

“…For example, the applied collision energy has a profound impact on the information content of MS/MS spectra [46]. Thus the optimal collision energies need to be determined for accurate fragment ion intensity predictions [36]. In addition, because of the distinct spectral characteristics of immunopeptides, fragment ion intensity prediction tools should be retrained using immunopeptidomics data to drastically improve prediction accuracy [12, 35].…”