Carolina Barra

Antigen-processing rewiring expose cryptic self promoting organ-specific autoimmunity

The paper identifies two distinct mechanisms of autoimmunity: altered antigen processing exposing cryptic self-antigens and breakdown of tolerance toward normally presented self-antigens.

Chemically informed representations of amino acids enable learning beyond the canonical protein alphabet

The paper introduces chemically informed peptide representations to enable machine learning models to learn from molecular structure rather than canonical amino acid symbols, improving prediction of MHC class I binding.

Resolving Genome-to-Phenotype Links in Bacteria: Machine-Learned Inference from Downsampled k-mer Representations

This paper introduces a downsampling method for bacterial genomes using k-mer representations, showing that ensemble models outperform deep learning on phenotype prediction tasks.

Multiple instance learning on tile level-pathologyimages provides accurate and interpretableclassification for breast cancer molecular subtypes

This paper presents a weakly supervised deep learning approach for breast cancer molecular subtyping using tile-level pathology images without tumor annotations, achieving high accuracy and interpretability.

ProtParts, an automated web server for clustering and partitioning protein datasets

ProtParts clusters protein sequences using BLAST E-values and graph algorithms to create independent partitions, improving model performance by reducing overfitting.

Integration of HLA-DR linkagedisequilibrium to MHC class II predictions

The paper introduces HLAAssoc-1.0 to infer HLA-DRB3/4/5 alleles from HLA-DRB1 data, improving allele and motif deconvolutions in immunopetidomics and training of binding prediction models.

Machine learning reveals limited contribution of trans-only encoded variants to the HLA-DQ immunopeptidome by accurate and comprehensive HLA-DQ antigen presentation prediction

The paper improves HLA-DQ antigen presentation prediction using machine learning and shows trans-only variants have limited impact on the immunopeptidome.

NetAllergen, a random forest model integrating MHC-II presentation propensity for improved allergenicity prediction

NetAllergen improves allergenicity prediction by integrating MHC-II presentation propensity, outperforming existing methods when sequence similarity to known allergens is low.

Improved prediction of MHC II antigen presentation throughintegration and motif deconvolution of mass spectrometryMHC eluted ligand data

The paper improves MHC II epitope prediction by integrating mass spectrometry data with a new machine learning framework, achieving better performance than existing methods.

NNAlign_MA; MHC peptidome deconvolution for accurate MHC binding motif characterization and improved T cell epitope predictions

NNAlign_MA deconvolves poly-specific MHC peptidome data to accurately characterize binding motifs and improve T cell epitope predictions.

Computational Tools for the Identification and Interpretation of Sequence Motifs in Immunopeptidomes

The paper describes computational approaches for analyzing immunopeptidomics data, focusing on motif identification, false hit filtering, and machine learning applications for predicting MHC ligands.