Biology subjects

Schumacher, T.

Publications and source records attributed to Schumacher, T..

3 recordsLinked to original sources

bioRxiv2023

Gene and protein sequence features augment HLA class I ligand predictions

The sensitivity of malignant tissues to T cell-based cancer immunotherapies is dependent on the presence of targetable HLA class I ligands on the tumor cell surface. Peptide intrinsic factors, such as HLA class I affinity, likelihood of proteasomal processing, and transport into the ER lumen have all been established as determinants of HLA ligand presentation. However, the role of sequence features at the gene and protein level as determinants of epitope presentation has not been systematically evaluated. To address this, we performed HLA ligandome mass spectrometry on patient-derived melanoma lines and used this data-set to evaluate the contribution of 7,124 gene and protein sequence features to HLA sampling. This analysis reveals that a number of predicted modifiers of mRNA and protein abundance and turn-over, including predicted mRNA methylation and protein ubiquitination sites, inform on the presence of HLA ligands. Importantly, integration of gene and protein sequence features into a machine learning approach augments HLA ligand predictions to a comparable degree as predictive models that include experimental measures of gene expression. Our study highlights the value of gene and protein features to HLA ligand predictions.

immunology↗

bioRxiv2023

Lack of detectable neoantigen depletion in treatment-naive cancers

While neoantigen depletion, a form of immunoediting due to Darwinian pressure exerted by the T cell based immune system during tumor evolution, has been clearly described in murine models, its prevalence in treatment-naive, developing human tumors remains controversial. We developed two novel methodologies to test for depletion of predicted neoantigens in patient cohorts, which both compare patients in terms of their expected number of neoantigens per mutational event. Application of these strategies to TCGA patient cohorts showed that neither basic nor more extensive versions of the methodologies, controlling for confounding factors such as genomic loss of the HLA locus, provided statistically significant evidence for neoantigen depletion. In the subset of analyses that did show a trend towards neoantigen depletion, statistical significance was not reached and depletion was not consistently observed across HLA alleles. Our results challenge the notion that neoantigen depletion is detectable in cohorts of unmatched patient samples using HLA binding prediction-based methodology.

cancer biology↗

bioRxiv2023

STAPLER: Efficient learning of TCR-peptide specificity prediction from full-length TCR-peptide data

The prediction of peptide-MHC (pMHC) recognition by {beta} T-cell receptors (TCRs) remains a major biomedical challenge. Here, we develop STAPLER (Shared TCR And Peptide Language bidirectional Encoder Representations from transformers), a transformer language model that uses a joint TCR{beta}- peptide input to allow the learning of patterns within and between TCR{beta} and peptide sequences that encode recognition. First, we demonstrate how data leakage during negative data generation can confound performance estimates of neural network-based models in predicting TCR - pMHC specificity. We then demonstrate that, because of its pre-training and fine-tuning masked language modeling tasks, STAPLER outperforms both neural network-based and distance-based ML models in predicting the recognition of known antigens in an independent dataset, in particular for antigens for which little related data is available. Based on this ability to efficiently learn from limited labeled TCR- peptide data, STAPLER is well-suited to utilize growing TCR - pMHC datasets to achieve accurate prediction of TCR - pMHC specificity.

bioinformatics↗