Pipeline release! nf-core/mhcquant v3.1.0 - 3.1.0 - BlüBa!
Identify and quantify MHC eluted peptides from mass spectrometry raw data
Please see the changelog: https://github.com/nf-core/mhcquant/releases/tag/3.1.0
#dda #immunopeptidomics #massspectrometry #mhc #openms #peptides #nfcore #openscience #nextflow #bioinformatics
Pipeline release! nf-core/mhcquant v3.0.0 - MHCquant 3.0.0 - Monrepos!
Please see the changelog: https://github.com/nf-core/mhcquant/releases/tag/3.0.0
#dda #immunopeptidomics #massspectrometry #mhc #openms #peptides #nfcore #openscience #nextflow #bioinformatics
Pipeline release! nf-core/mhcquant v2.6.0 - MHCquant 2.6.0 - Mr Bob!
Please see the changelog: https://github.com/nf-core/mhcquant/releases/tag/2.6.0
#dda #immunopeptidomics #mass-spectrometry #mhc #openms #peptides #nfcore #openscience #nextflow #bioinformatics
From these MS/MS spectrum and PSM characteristics, can you guess whether it's #immunopeptidomics or #BottomUpProteomics data?
Join us in the #BioinformaticsHub at #HUPO2023 today to discuss #MachineLearning applications for immunopeptidomics data analysis.
Slides: https://doi.org/10.5281/zenodo.8353779
Notebook to generate the figures: https://gist.github.com/bittremieux/af2985306c4121ab70ac66f44398951c
Immunopeptidomics data analysis is very challenging:
- A massive search space has to be considered (all subsequences of the proteins under consideration).
- Immunopeptides have different characteristics than standard tryptic peptides (non-tryptic peptide termini, predominantly singly charged).
This leads to low spectrum annotation rates in #immunopeptidomics. 😥
Our work on fragment ion intensity prediction for the analysis of timsTOF #immunopeptidomics data is now online on #bioRxiv: https://www.biorxiv.org/content/10.1101/2023.07.17.549401v1
Check out this great effort by Charlotte Adams, who basically learned #DeepLearning/#MachineLearning from scratch for this project.
Thread. 👇
The slides from my presentation at the #BSPREUPA2023 conference on timsTOF fragment ion intensity prediction for #immunopeptidomics are available here: https://doi.org/10.5281/zenodo.8163554
Fantastic work by Charlotte Adams. Stay tuned for the full preprint very soon.
Introduction Immunopeptidomics plays a crucial role in identifying targets for immunotherapy and vaccine development. Because immunopeptides are generated from their parent proteins in an unpredictable manner, rather than being able to use known digestion rules, every possible protein subsequence needs to be considered. This leads to an inflation of the search space and results in a low spectrum identification rate. Rescoring is a powerful enhancement of standard database searching that boosts the spectrum identification sensitivity and accuracy by unlocking the intensity dimension of MS/MS spectra with peptide fragment intensity predictions. The high sensitivity of a timsTOF instrument makes it ideal for detecting immunopeptides that are present at relatively low abundances. To improve the identification rate and the reliability of immunopeptidomics experiments performed using timsTOF instruments, we developed an optimized fragment ion intensity prediction model based on Prosit. Methods We analyzed over 300,000 synthesized non-tryptic peptides from the ProteomeTools project on a TimsTOF-Pro (Bruker, Bremen). The spectra were searched using MaxQuant (version 2.1.2.0) at 1% PSM-level FDR. The 277,781 obtained PSMs (93,227 non-tryptic measured in this study and 184,554 previously published tryptic peptide PSMs) were split into training (80%), validation (10%), and test (10%) sets. The training set was used to fine-tune the existing Prosit fragment intensity prediction model, the validation set to control for overfitting with early stopping, and the test set to evaluate the model. We reprocessed immunopeptidomics timsTOF data from a recent study using MaxQuant (version 2.0.3.1) and rescored all proposed PSMs by integrating the fragment intensity predictions. Results Comparison of the previously published and the here developed Prosit models showed a substantially improved normalized spectral contrast angle (SA) between predicted and experimental spectra for non-tryptic peptides (SA ≥ 0.9 for 2.4% vs 26.3% of spectra, respectively) and for tryptic peptides (SA ≥ 0.9 for 0.2% vs 42.1%). To evaluate whether rescoring with our model is able to increase the identification rate we reprocessed public HLA Class I and Class II immunopeptidome data. Similarly to what was observed previously on Orbitrap instruments, incorporating our model into the database matching process increased the spectrum identification rate of immunopeptides measured on a timsTOF. Compared to MaxQuant, we identified 3.0-fold more HLA class I peptides and 1.7-fold more HLA class II peptides after rescoring. To evaluate the clinical relevance of rescoring, we will look for peptides exclusively presented by tumors. We hypothesize that rescoring results in an increased reliability and identification rate of neoepitopes.
nf-core
Wout Bittremieux