Lukas Käll

borgonovo improves protein sequence accuracy from tandem mass spectra by consensus assembly of redundant de novo peptide reads, achieving 0.87 mean fixed-window identity on therapeutic antibodies.

DORSSAA provides a platform for analyzing stability and solubility alteration data to identify drug-protein interactions and therapeutic targets.

This paper introduces Pairwise Attention, a modified transformer encoder for de novo peptide sequencing from MS2 spectra, improving precision by 12.7% on a nine-species benchmark.

Met-ID is an open-source software for annotating chemical modifications in MALDI-MSI data, combining MS1 and MS2 functionalities with a customizable database.

The paper introduces Percolator-RESET, a post-processor for tandem mass spectrometry proteomics that better controls the false discovery rate than Percolator.

The paper evaluates using cryoEM and mass spectrometry with a deep-learning tool to sequence antibodies and map epitopes from convalescent serum, showing 80-90% accuracy in variable domain sequencing.

This paper introduces a deep learning method for spatial landmark detection and tissue registration, showing improved accuracy and stability over existing approaches.

ModelAngelo combines convolutional and graph neural networks with protein language model embeddings and hidden Markov model searches to automate atomic model building and identify unknown proteins in cryo-EM maps.

The paper introduces MIPath, a pathway analysis method using mutual information to assess pathway-phenotype associations with fewer assumptions, showing better performance on single-cell datasets.

ProHap constructs protein sequence databases from phased genotypes to account for haplotypic diversity in proteomic searches.

This paper introduces a spatial multimodal analysis method that integrates histology, mass spectrometry imaging, and spatial transcriptomics to study mRNA and metabolite distributions in brain tissues.

This study improves variant peptide identification in proteomics by integrating retention time and fragmentation predictors into a proteogenomic pipeline, enhancing discrimination between correct and incorrect matches.

The paper investigates how common genetic haplotypes affect proteomic search results and identifies multi-variant peptides detectable by mass spectrometry.

The paper evaluates consensus spectrum generation methods in proteomics, finding BEST and BIN to be the most reliable for peptide identification.

The paper uses pathway activity estimated via singular value decomposition to predict survival time in breast cancer patients, finding it more prognostic than individual transcripts.

The paper introduces Triqler, a Bayesian model that integrates quantification and identification uncertainties for differential protein abundance analysis in shotgun proteomics.

The paper introduces a GPU-accelerated parallelization of the Green algorithm for permutation tests, significantly reducing computation time for medium sample sizes.

The paper introduces Quandenser, a method for shotgun proteomics that clusters quantification data to improve differential expression analysis by reducing noise and increasing sensitivity.

The paper develops a computational tool to analyze protein complex behavior across human cell types using mass spectrometry data and provides a web service for co-regulation analysis.

The paper introduces Triqler, a probabilistic graphical model for shotgun proteomics that controls FDR and improves sensitivity in protein quantification.

The paper introduces a protein standard with shared peptides to benchmark protein inference algorithms, showing that methods excluding shared peptides provide more accurate error estimates.