Robert Hoehndorf

PhysioMap connects physiological traits to causal relationships and predicts responses in metabolic diseases, showing agreement with references after excluding discordant cases.

This study identifies miRNA-regulated networks in human heatstroke that link stress-survival signalling to impaired protein quality control and metabolic flexibility.

This paper reports on the fourth CAFA challenge, showing improvements in Gene Ontology term prediction but highlighting challenges in new evaluation settings.

PAVS is a database of phenotype-associated variants from Saudi Arabian rare disease patients, enabling gene prioritization through phenotype-based similarity search.

METEOR integrates enzyme prediction with genome-scale metabolic model reconstruction to improve functional annotations and network viability.

This study identifies a novel mechanism of phase separation driven by the Janus-class KFL region of PYK2, showing its role in kinase activation and cell adhesion.

This study uses whole-genome sequencing to analyze S. aureus isolates from Saudi Arabia, revealing higher genetic diversity and antimicrobial resistance in pilgrimage cities compared to non-pilgrimage cities.

The paper constructs a pangenome graph reference for Japanese and Saudi Arabian populations, showing improved variant detection compared to linear references and comparable performance to existing pangenome graphs.

NanoDesigner improves de novo nanobody design success rates by iteratively refining docking and CDR generation using an expectation maximization algorithm.

The paper evaluates genome-scale protein function annotations for completeness, coherence, and consistency, finding that computational methods fail to produce biologically plausible annotations despite high individual protein accuracy.

This paper introduces an inductive method to predict gene-disease associations using phenotype ontologies, achieving performance comparable to transductive methods while generalizing to novel diseases.

The paper develops a transformer-based model to predict protein functions in bacterial genomes using genomic context, showing that synteny significantly improves prediction performance.

This paper introduces PU-GO, a method for protein function prediction that addresses the positive-unlabeled ranking problem using ontology-based priors, showing improved performance on benchmark datasets.

This paper introduces EmbedPVP, a method that prioritizes genomic variants in genetic diseases by integrating genomic data with clinical phenotypes and background knowledge.

The paper evaluates variant prioritization methods for rare disease diagnosis using data from the Rare Genomes Project, finding that top models identify causal variants in most solved families and uncover potential candidates in unsolved ones.

DeepGOMeta predicts protein functions for microbes using a deep learning model trained on microbial data, improving function prediction for novel proteins.

DeepGO-SE improves protein function prediction for distantly related proteins by using a pretrained large language model and a neuro-symbolic model that leverages GO axioms.

LEP-AD uses a Transformer protein language model to predict drug-target interactions, achieving state-of-the-art results across multiple datasets.

BORD improves biomedical concept recognition by using distant supervision and ontology hierarchy, outperforming existing methods for disease and phenotype identification.

The paper presents a fully annotated reference genome from a Saudi individual using de novo assembly and AI-based annotation, making all data publicly available.

FrameRate predicts coding frames from unassembled metagenomic reads using machine learning, bypassing assembly and capturing functional profiles efficiently.

The paper assesses how model organism phenotypes contribute to identifying human disease genes and finds that only mouse phenotypes accurately predict human gene-disease associations.

DeepGOZero improves protein function prediction by using ontology axioms and zero-shot learning to predict functions with few or no annotations.

DeepSVP prioritizes pathogenic structural variants by integrating genotype and phenotype data using deep learning, improving benchmark performance and identifying novel variants in consanguineous families.

DTI-Voodoo predicts drug-target interactions using molecular features, phenotypic effects, and protein-protein interaction networks, showing improved performance over existing methods.

Klarigi identifies compositional and discriminatory ontology classes for groups of biomedical entities, providing semantic explanations that extend existing knowledge.

The paper expands ontology labels to improve clinical text characterization by linking synonyms across ontologies, enhancing semantic similarity for patient annotation.

This paper develops a method to use uncurated clinical text phenotypes for differential diagnosis of common diseases using semantic similarity.

DeepMOCCA integrates multi-omics data with biological networks to predict cancer survival and identify patient-specific prognostic markers.

DeepViral improves prediction of virus-host protein interactions by incorporating disease phenotypes and protein sequences.