Robbin Bouwmeester

The paper evaluates non-linear models for peptide-spectrum match rescoring in proteomics, finding that complex models increase PSMs but introduce biased FDR estimates.

The paper introduces a deep learning model to predict peptide retention times in liquid chromatography and identifies key factors influencing retention through saliency mapping.

The paper generates a benchmark dataset to evaluate proteomics instruments and workflows using high-throughput LC gradients across multiple platforms and organisms.

The paper presents the E. coli PeptideAtlas build, a comprehensive pan-proteome resource covering 4,755 proteins and over 10,000 PTM sites from 40 datasets.

quantms-rescoring improves proteomics analysis by integrating multiple search engines and machine learning, increasing identified spectra and quantifying more phosphorylated peptides and modification sites.

iDeepLC improves retention time prediction for peptides with unseen modifications by incorporating chemical structural information, outperforming previous models.

ProteoBench provides a centralized platform for comparing proteomics data analysis workflows, enabling benchmarking and community collaboration.

This study benchmarks eight de novo peptide sequencing models on proteomics datasets to evaluate their performance and identify limitations in resolving sequence ambiguity.

The paper improves IM2Deep to predict collisional cross-sections for peptides with multiple conformations, using a comprehensive dataset and tailored architecture.

DeepLC uses transfer learning to improve LC retention time prediction across different experimental setups and peptide modifications.

The paper improves bacterial epitope discovery by integrating multiple search engines and rescoring, increasing peptide identification by 27% and detecting 18 new bacterial peptides.

ClassiCOL improves species classification from ancient collagen LC-MS/MS data by resolving peptide ambiguity through isoBLAST and taxonomic branching analysis.

TIMS2Rescore is a data-driven rescoring pipeline optimized for DDA-PASEF data that improves protein identification in proteomics analyses.

The paper reanalyzes human tissue proteomics data using pangenome references to identify novel peptides and variants, demonstrating the effectiveness of proteogenomics approaches with pangenome data.

FAVA infers high-quality functional protein interaction networks from scRNA-seq and proteomics data, outperforming existing methods and identifying interactions for understudied proteins.

This study benchmarks 12 DIA analysis workflows for shotgun proteomics samples, finding DIA-NN and Spectronaut in library-free mode offer highest sensitivity with good reproducibility and accuracy.

A machine learning model trained on proteomics data from 183 human datasets identifies tissue and cell-type specific proteins with high accuracy.

This study evaluates PTM predictors using protein language models, revealing a training bias from proteomics data that overestimates prediction accuracy and suggests a solution through negative sample filtering.

MS2Rescore improves immunopeptide identification rates by integrating retrained MS2PIP, DeepLC, and Percolator, achieving higher spectrum identification rates and unique peptide counts.

The paper develops SECRiFY, a method to assess secretability of protein fragments in yeast, revealing intrinsic disorder and flexibility as key factors.

DeepLC predicts retention times for modified peptides, improving identification accuracy in LC-MS workflows by handling unseen modifications.

ionbot improves peptide identification in mass spectrometry by integrating machine learning and open modification searches, increasing PSM confidence and identification rates.

The paper compares proteogenomics and open variant search approaches for detecting single amino acid variants in shotgun proteomics data using NA12878 cell line data.

The paper evaluates integrating Percolator into COSS for improved peptide identification sensitivity and specificity using multiple datasets and search tools.

The paper introduces generalized calibration to map retention time models between different liquid chromatography setups, improving prediction accuracy across setups.

The paper shows that using MS2PIP to predict fragment ion intensities and comparing them to experimental spectra improves Percolator's ability to distinguish PSMs while controlling false discovery rates.

This study evaluates seven machine learning algorithms for predicting LC retention times across 36 metabolomics datasets, finding no single algorithm is optimal for all cases and highlighting the benefits of combining models.