Fragment ion intensity prediction improves the identification rate of non-tryptic peptides in timsTOF

Abstract

Immunopeptidomics is crucial for immunotherapy and vaccine development. Because the generation of immunopeptides from their parent proteins does not adhere to clear-cut rules, rather than being able to use known digestion patterns, every possible protein subsequence within human leukocyte antigen (HLA) class-specific length restrictions needs to be considered during sequence database searching. This leads to an inflation of the search space and results in lower spectrum annotation rates. Peptide-spectrum match (PSM) rescoring is a powerful enhancement of standard searching that boosts the spectrum annotation performance. We analyze 302,105 unique synthesized non-tryptic peptides from the ProteomeTools project on a timsTOF-Pro to generate a ground-truth dataset containing 93,227 MS/MS spectra of 74,847 unique peptides, that is used to fine-tune the deep learning-based fragment ion intensity prediction model Prosit. We demonstrate up to 3-fold improvement in the identification of immunopeptides, as well as increased detection of immunopeptides from low input samples.

Machine Learning Strategies to Tackle Data Challenges in Mass Spectrometry-Based Proteomics

“…23 The same peptides were also measured using an Evosep One HPLC system (Evosep) coupled to a hybrid TIMS-quadrupole TOF mass spectrometer (Bruker Daltonik timsTOF Pro). 33 ProteomeTools also includes a TMT subset, containing both tryptic and nontryptic synthesized peptides labeled with a TMT 6-plex label. Peptide pools were subjected to LC using a Dionex 3000 HPLC system (Thermo Fisher Scientific) with C18 columns coupled to an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific), after which each pool was measured twice, covering seven different fragmentation modes: "TMT1" and "TMT2".…”

“…Each peptide pool was first measured using a “survey” run, followed by three subsequent LC–MS/MS measurements comprising a total of 11 different fragmentation modes: 3xHCD, 2xIT_2xHCD, and ETD, covering different fragmentation settings (HCD, CID, ETD, EThcD, and ETciD) and mass analyzers (FTMS: Orbitrap mass analyzer, ITMS: ion trap mass analyzer) . The same peptides were also measured using an Evosep One HPLC system (Evosep) coupled to a hybrid TIMS-quadrupole TOF mass spectrometer (Bruker Daltonik timsTOF Pro) …”

“…Transfer learning is another related approach that we briefly touched upon, which has shown promise in various domains. ,− These methods essentially combine pretraining and multitask learning, enabling the transfer of knowledge from one task to another. As an example, transfer learning has enabled the adaptation of models like Prosit to predict fragment ion intensities for different types of MS data …”

“…As an example, transfer learning has enabled the adaptation of models like Prosit to predict fragment ion intensities for different types of MS data. 33 Finally, we would like to call upon the proteomics community to prioritize the generation, compilation, and curation of data sets. While the development of new ML models is often more highlighted, the community stands to benefit more substantially from robust data set creation and rigorous benchmarking.…”

Investigating Antiprotozoal Chemotherapies with Novel Proteomic Tools—Chances and Limitations: A Critical Review

“…Efforts are continuously made to boost the number of validated peptides and proteins by including further features in the estimation of FDR [42] and rescoring in specific application contexts [43]. Since correct proteins are often identified by multiple peptides and incorrect proteins by one random match only, a good practice is to require at least two unique peptides to consider protein X as "identified" [44].…”