Biomedical Text Mining and Ontologies

FuncFetch: An LLM-assisted workflow enables mining thousands of enzyme-substrate interactions from published manuscripts

FuncFetch uses an LLM to mine enzyme-substrate interactions from manuscripts, achieving high precision and recall in extracting functional data.

VaxLLM: Leveraging Fine-tuned Large Language Model for automated annotation of Brucella Vaccines

VaxLLM automates Brucella vaccine article classification and annotation with high accuracy, outperforming non-fine-tuned models and enabling structured data integration.

The effectiveness of Large Language Models with RAG for auto-annotating phenotype descriptions

This study uses Large Language Models with Retrieval-Augmented Generation to auto-annotate plant phenotypes with ontology terms, achieving results comparable to expert annotations.

Annotating publicly-available samples and studies using interpretable modeling of unstructured metadata

This paper introduces txt2onto 2.0, a method for annotating biomedical metadata with disease and tissue terms using interpretable machine learning, showing improved performance over its predecessor.

Network for Knowledge Organization (NEKO): an AI knowledge mining workflow for synthetic biology research

NEKO is an AI workflow using LLMs to extract and organize scientific knowledge from literature, providing more specific and actionable insights than existing models.

Generating Biomedical Knowledge Graphs from Knowledge Bases, Registries, and Multiomic Data

The paper describes the transformation of multiomics and EHR data into privacy-preserving biomedical knowledge graphs within the Biomedical Data Translator ecosystem.

A machine learning framework for extracting information from biological pathway images in the literature

This paper introduces a machine learning framework that extracts biological pathway information from literature images and compares it to manually curated maps.

How do Large Language Models understand Genes and Cells

This study evaluates seven large language models on their ability to understand genes and cells, finding they have some capability but require improvements in cell annotation tasks.

Automated Harmonization and Large-Scale Integration of Heterogeneous Biomedical Sample Metadata Using Large Language Models

This paper introduces a framework using large language models to automate metadata harmonization and integration across biomedical studies, achieving higher recovery rates than traditional methods.

Harnessing GPT-4 for Automated Curation of E3-Substrate Relationships in the Ubiquitin-Proteasome System

This paper uses GPT-4 to automate curation of E3-substrate relationships from literature, resulting in a database of 3,294 pairs with high accuracy.

Lifestyle factors in the biomedical literature: An ontology and comprehensive resources for named entity recognition

The paper develops a Lifestyle Factor Ontology and annotated corpus for detecting lifestyle factors in biomedical texts, achieving 76% F-score with a transformer-based NER system.

Knowledge Graph-based Thought: a knowledge graph enhanced LLMs framework for pan-cancer question answering

The paper introduces a framework that combines large language models with knowledge graphs to improve factual accuracy in biomedical question answering, demonstrating effectiveness in drug repositioning and resistance prediction.

Building a literature knowledge base towards transparent biomedical AI

The paper develops a knowledge graph framework to extract biomedical knowledge from PubMed, creating a database that enhances transparency in biomedical AI applications.

Beyond associations: A benchmark Causal Relation Extraction Dataset (CRED) of disease-causing genes, its comparative evaluation, interpretation and application

The paper introduces a benchmark dataset (CRED) for causal relation extraction from biomedical literature, evaluates classifiers, and applies it to identify Parkinson's disease-causing genes.

Mining impactful discoveries from the biomedical literature

This paper introduces a method to mine past discoveries from the biomedical literature by detecting surges in relation prevalence over time, providing a benchmark dataset for Literature-Based Discovery evaluation.

Ayurveda based phenotype annotation integrated with Expectation Maximization algorithm reveal six biological clusters of rare diseases: potential in resolving ciliopathies

This study integrates Ayurveda with modern medical data to classify rare diseases into six clusters, revealing potential ciliopathies in specific clusters.

A natural language processing system for the efficient extraction of cell markers

This study develops MarkerGeneBERT, an NLP system that extracts cell markers from scRNA-seq literature, achieving 76% completeness and 75% accuracy in marker gene identification.

STRING-ing together protein complexes: corpus and methods for extracting physical protein interactions from the biomedical literature

The paper introduces a manually annotated corpus and a relation extraction model to identify physical protein interactions from biomedical literature, integrating results into the STRING database.

FoodAtlas: Automated Knowledge Extraction of Food and Chemicals from Literature

The paper develops a pipeline using large language models to extract food-chemical relationships from literature, identifying many novel interactions.

Leveraging Generative AI to Accelerate Biocuration of Medical Actions for Rare Disease

AutoMAxO uses LLMs to semi-automate MAxO curation for rare diseases, identifying 538 confirmed disease-treatment annotations from 4,918 medical abstracts.

VieMedKG: Knowledge Graph and Benchmark for Traditional Vietnamese Medicine

This paper constructs the first knowledge graph for Traditional Vietnamese Medicine (TVM) by adapting existing Traditional Chinese Medicine (TCM) resources and develops a benchmark for knowledge graph-based question-answering.

LitGene: a transformer-based model that uses contrastive learning to integrate textual information into gene representations

LitGene integrates textual information into gene representations using a transformer-based model with contrastive learning, improving disease risk gene prediction and interpretability.

VIOLIN: A modular framework for scalable reconciliation of heterogeneous interaction graphs

VIOLIN reconciles heterogeneous interaction graphs with extracted molecular interactions, showing stable and interpretable results across different models and datasets.

Semantic computing for human phenotypes

This paper explores using natural language and ontologies to integrate genetic disease risk studies and identifies areas for improvement in data harmonization tools.

Improving dictionary-based named entity recognition with deep learning

The paper improves dictionary-based named entity recognition by using a Transformer model to automatically generate block lists, increasing precision in text mining.

Harnessing generative AI to annotate the severity of all phenotypic abnormalities within the Human Phenotype Ontology

This paper uses GPT-4 to automatically annotate the severity of phenotypic abnormalities in the Human Phenotype Ontology, achieving high accuracy in clinical characteristic identification.

Publication, Funding, and Experimental Data in Support of Human Reference Atlas Construction and Usage

This paper compiles a dataset linking HRA digital objects to experts, publications, funding, and datasets to support systematic HRA construction and usage.

Iterative Prompt Refinement for Mining Gene Relationships from ChatGPT

This study improves gene relationship prediction using iterative prompt refinement with ChatGPT, achieving better performance for complex gene interactions.

Fine-tuning Large Language Models for Rare Disease Concept Normalization

This paper fine-tunes Llama 2 for rare disease concept normalization using HPO data, achieving high accuracy with synonym inclusion and typo handling.

Semantic harmonization of Alzheimer's disease datasets using AD-Mapper

This paper introduces AD-Mapper, a tool for semantic harmonization of Alzheimer's disease datasets, which outperforms a string-matching baseline.