David Gfeller

This paper evaluates AlphaFold3's performance across diverse structural biology tasks, finding uneven accuracy and strong dependence on training data.

This study shows autophagy contributes to MHC class II presentation of RA-related peptides in HLA-DRB1*04:01 alleles, with 30% of peptides differing between autophagy sufficient and deficient cells.

This study shows that unpaired TCRα + TCRβ sequencing can train accurate TCR-epitope recognition predictors at lower cost than paired sequencing.

SuperCell2.0 constructs multimodal metacell atlases from blood and tumor samples, identifying interferon-primed monocytes and macrophages with improved inter-modality consistency.

The paper quantifies how V and J gene usage affects CDR3 length and amino acid composition in TCR repertoires and provides a quality control framework for TCR sequencing data.

The paper introduces TCR specificity profiles to identify key determinants of TCR-epitope recognition and shows their ability to predict cross-reactivity and interpret machine learning models.

The paper uses phage display and machine learning to discover TCRs specific to cancer antigens, identifying functional TCRs from large repertoires.

This study evaluates how well MHC-I ligand predictions can be extrapolated across alleles and species using an improved predictor, finding high accuracy in human and mouse but lower in other species.

SuperSpot reduces spatial transcriptomic data size and noise by grouping similar spots into metaspots using graph-based clustering.

This paper develops EPIC-ATAC to estimate cancer and immune cell-type proportions from bulk tumor ATAC-Seq data, showing accurate prediction of immune contexture in breast cancer subtypes.

The paper evaluates deconvolution methods for inferring immune infiltration from bulk gene expression using a DREAM Challenge, finding that some community-contributed methods, especially deep learning-based ones, perform well.

The paper reviews metacell applications in single-cell genomics, provides a tutorial, and introduces a pipeline for metacell construction and analysis.

This paper introduces MixTCRpred, a deep learning model that predicts TCR-epitope interactions and identifies epitope-specific T cells, showing enrichment in COVID-19 patients.

This study introduces STACAS v2, a semi-supervised method for integrating scRNA-seq data that outperforms existing methods by preserving biological variability using cell type annotations.

This paper describes using MixMHC2pred to predict MHC-II binding specificity and ligands from amino acid sequences across species.

The paper introduces metacells to address scalability issues in scRNA-seq data by merging similar cells, improving downstream analyses and accelerating atlas construction.

The paper improves epitope prediction by integrating HLA-I ligand data and TCR recognition models, identifying potent SARS-CoV-2 CD8+ T-cell epitopes with experimental validation.

The paper reanalyzes the contribution of proteasomal spliced peptides to the HLA-I ligandome and finds that only 2-4% of the ligandome may be explained by such peptides.

The paper identifies HLA-I binding motifs from peptidomics data to improve neo-antigen predictions and reveals allosteric regulation of HLA specificity.