Roman Joeres

This study identifies information leakage in enzyme-substrate prediction models and datasets, showing performance drops when leakage is removed.

ALPAR automates AMR data analysis from genomic data processing to ML model training and GWAS, offering a reproducible pipeline for single-reference AMR studies.

StructGuy predicts functional effects of genetic variants using a data leakage-free supervised approach, achieving competitive performance against zero-shot methods.

This paper introduces EMMA, a transformer-based multi-task learning framework for predicting enzyme-small molecule interactions, showing strong performance across different data-splitting regimes.

GlyContact analyzes glycan 3D structures from MD, NMR, or X-ray data to understand structure-function relationships and improve AI models for lectin-glycan binding.

This paper introduces a framework for automated glycan nomenclature conversion that accepts any input format and outputs a standardized IUPAC-condensed format, enabling seamless integration of glycan data.

This paper develops a machine learning workflow to identify generalizable diagnostic ions for O-glycan isomer differentiation using a curated dataset of 237,000 spectra.

DataSAIL splits biological datasets to minimize information leakage by reducing similarity between splits, improving model generalization assessment.

GlyLES converts IUPAC-condensed glycan notation to SMILES using a grammar-based approach, enabling atomic-level representation and applications in molecular modeling and machine learning.