A comprehensive LFQ benchmark dataset on modern day acquisition strategies in proteomics

Abstract

In the last decade, a revolution in liquid chromatography-mass spectrometry (LC-MS) based proteomics was unfolded with the introduction of dozens of novel instruments that incorporate additional data dimensions through innovative acquisition methodologies, in turn inspiring specialized data analysis pipelines. Simultaneously, a growing number of proteomics datasets have been made publicly available through data repositories such as ProteomeXchange, Zenodo and Skyline Panorama. However, developing algorithms to mine this data and assessing the performance on different platforms is currently hampered by the lack of a single benchmark experimental design. Therefore, we acquired a hybrid proteome mixture on different instrument platforms and in all currently available families of data acquisition. Here, we present a comprehensive Data-Dependent and Data-Independent Acquisition (DDA/DIA) dataset acquired using several of the most commonly used current day instrumental platforms. The dataset consists of over 700 LC-MS runs, including adequate replicates allowing robust statistics and covering over nearly 10 different data formats, including scanning quadrupole and ion mobility enabled acquisitions. Datasets are available via ProteomeXchange (PXD028735).

ProteomicsML: An Online Platform for Community-Curated Data sets and Tutorials for Machine Learning in Proteomics

“…20 Lately the field has published multiple ML and DL approaches for both peptide and metabolite CCS prediction. [21][22][23] The tutorials made available in ProteomicsML use both TIMS and TWIMS data, where the large TIMS data set is from Meiers et al 23 (718,917 data points) and the TWIMS data is from Puyvelde et al 24 (6268 data points). The tutorial is a walkthrough that trains linear models to more complex non-linear models (e.g., DL based networks) showing advantages and disadvantages of the learning algorithms for CCS prediction.…”

Integration of data‐independent acquisition (DIA) with co‐fractionation mass spectrometry (CF‐MS) to enhance interactome mapping capabilities

“…Direct comparisons between library-based and library-free approaches have shown that library-free approaches not only identify a higher number of peptides and proteins for a given experiment, but they also take less time and resources overall [2,69]. The recently published benchmarking dataset would also be a strong dataset for investigating the overall performance of library-based versus library-free approaches [55]. Furthermore, hybrid approaches, where DDA and DIA are combined into the same run, such as mixed-data acquisition [70] or data-dependent-independent acquisition [71], removes the requirement for generating separate spectral libraries.…”

Quality control for the target decoy approach for peptide identification

“…MS runs from three publically available datasets were downloaded from the ProteomePro-teomeXchange Consortium (http://proteomecentral.proteomexchange.org) PRIDE partner repository 5 : A Pyrococcus furiosus dataset, a human dataset, and an immunopeptidomics dataset, with identifiers PXD001077, PXD028735, and PXD021398, respectively [6][7][8] . Raw MS datasets were fully reanalyzed, except for the immunopeptidomics dataset, where publically available search results were used.…”