Biomedical Text Mining and Ontologies

The paper extracts pathogen-disease associations from literature using ontology and statistical methods, resulting in 3,420 associations integrated into a database.

The paper demonstrates that citizen scientists can perform relationship extraction from biomedical abstracts using Mark2Cure, highlighting the importance of accurate named entity recognition.

WikiMedMap maps phenotype strings to clinical vocabularies using Wikipedia, outperforming existing tools in matching accuracy.

The paper uses distant supervision to augment small manually annotated datasets for relation extraction, improving model performance after filtering noisy data.

The paper develops a graph-based framework to infer direct associations between HPO and GO terms, improving the interpretation of molecular processes in disease contexts.

The paper evaluates methods to improve generalization of deep learning models for biomedical named entity recognition (BioNER) and introduces an open-source tool for the task.

CoCoScore improves relation extraction by scoring co-occurrence with contextual awareness, outperforming existing methods in disease-gene, tissue-gene, and protein-protein association identification.

variant2literature searches biomedical literature for genetic variants by processing images and normalizing variant formats to improve recall.

PathoPhenoDB links human pathogens to disease phenotypes to support infectious disease research by integrating pathogen-disease associations and drug resistance information.

This paper addresses gaps in UMLS disease mappings by using alternative expert-curated sources, resolving over 50% of missing mappings for some vocabularies.

PPPred classifies protein-phenotype co-mentions from biomedical text using supervised learning, outperforming baselines in validation.

The paper creates a similarity corpus for microbial transcriptional regulation sentences, evaluated by curators to aid in biomedical literature curation.

CancerMine compiles text-mined cancer gene annotations and provides a resource for gene-centric cancer type clustering validated against TCGA data.

This paper uses temporally-enabled word embeddings to recapitulate and predict disease-gene associations from OMIM data, achieving high AUC and predicting some associations before their first literature mention.

The paper semantically integrates LOINC-encoded laboratory data with the Human Phenotype Ontology to identify asthma biomarkers from EHR data.

The paper develops a standardized ontology for emergency department presenting problems using real-world data and expert consensus, achieving high validation accuracy.

The paper introduces PGxO and PGxLOD to reconcile pharmacogenomic knowledge from diverse sources, enabling comparison of gene-drug-adverse event relationships.

The paper evaluates ontology design patterns for combining pathology and anatomy ontologies using an aging mice dataset, showing how different patterns perform in analysis tasks.

The paper introduces k-BOOM, a Bayesian method for merging ontologies by estimating logical relationships from mappings, applied to build a unified disease ontology.

The paper improves biomedical concept recognition using deep neural sequence models, achieving higher F1 scores for ontology concept identification in scientific texts.

BO-LSTM uses biomedical ontologies with LSTM networks to improve drug-drug interaction classification in text, achieving better performance with limited annotations.

The paper develops workflows to re-curate and enrich knowledge graphs encoded in Biological Expression Language using automated and manual methods.

This paper introduces a formal framework for representing biological mechanisms and applies it to analyze human genetic disease mechanisms.

ClinPhen automatically extracts and prioritizes patient phenotypes from medical records to improve genetic disease diagnosis efficiency.

Biochat links GEO datasets via metadata similarity to support multi-omics data integration through a web application.

Phenotero integrates ontology annotation into writing workflows by leveraging Zotero to reference ontology classes in clinical reports and manuscripts.

This paper introduces a system using deep reinforcement learning to automatically curate a genetic association database by simulating human research behaviors.

The paper introduces two NLP-based methods to measure semantic similarity of Gene Ontology terms using their definitions, achieving better classification accuracy than tree-based methods.

The paper introduces a multi-task learning framework for biomedical named entity recognition that improves performance by sharing information across entity types.

The paper identifies high-priority proteins across the human diseasome using semantic similarity and provides a systematic collection of popular proteins across diseases, phenotypes, and pathways.