Knowledge graph construction for heart failure using large language models with prompt engineering DOI Creative Commons

Tianhan Xu,

Yixun Gu, Mantian Xue

et al.

Frontiers in Computational Neuroscience, Journal Year: 2024, Volume and Issue: 18

Published: July 2, 2024

Constructing an accurate and comprehensive knowledge graph of specific diseases is critical for practical clinical disease diagnosis treatment, reasoning decision support, rehabilitation, health management. For construction tasks (such as named entity recognition, relation extraction), classical BERT-based methods require a large amount training data to ensure model performance. However, real-world medical annotation data, especially disease-specific samples, are very limited. In addition, existing models do not perform well in recognizing out-of-distribution entities relations that seen the phase.

Language: Английский

Towards electronic health record-based medical knowledge graph construction, completion, and applications: A literature study DOI Creative Commons
Lino Murali,

G. Gopakumar,

Daleesha M. Viswanathan

et al.

Journal of Biomedical Informatics, Journal Year: 2023, Volume and Issue: 143, P. 104403 - 104403

Published: May 24, 2023

Language: Английский

Citations

37
Overview of knowledge reasoning for knowledge graph DOI
Xinliang Liu,

Tingyu Mao,

Yanyan Shi

et al.

Neurocomputing, Journal Year: 2024, Volume and Issue: 585, P. 127571 - 127571

Published: April 1, 2024

Language: Английский

Citations

10
RT: a Retrieving and Chain-of-Thought framework for few-shot medical named entity recognition DOI Creative Commons
Mingchen Li, Huixue Zhou, Han Yang

et al.

Journal of the American Medical Informatics Association, Journal Year: 2024, Volume and Issue: 31(9), P. 1929 - 1938

Published: May 6, 2024

Abstract Objectives This article aims to enhance the performance of larger language models (LLMs) on few-shot biomedical named entity recognition (NER) task by developing a simple and effective method called Retrieving Chain-of-Thought (RT) framework evaluate improvement after applying RT framework. Materials Methods Given remarkable advancements in retrieval-based model across various natural processing tasks, we propose pioneering designed amalgamate both approaches. The approach encompasses dedicated modules for information retrieval processes. In module, discerns pertinent examples from demonstrations during instructional tuning each input sentence. Subsequently, module employs systematic reasoning process identify entities. We conducted comprehensive comparative analysis our against 16 other NER tasks BC5CDR NCBI corpora. Additionally, explored impacts negative samples, output formats, missing data performance. Results Our proposed outperforms LMs with micro-F1 scores 93.50 91.76 corpora, respectively. found that using positive (vs Tree-of-Thought) performed better. utilization partially annotated dataset has marginal effect Discussion is first investigation combine LLM methodology NER. aids retrieving most relevant sentence, offering crucial knowledge predict also meticulous examination methodology, incorporating an ablation study. Conclusion demonstrated state-of-the-art tasks.

Language: Английский

Citations

10
Drug-Target Interactions Prediction Based on Signed Heterogeneous Graph Neural Networks DOI Open Access
Ming Chen,

Yajian Jiang,

Xiujuan Lei

et al.

Chinese Journal of Electronics, Journal Year: 2024, Volume and Issue: 33(1), P. 231 - 244

Published: Jan. 1, 2024

Drug-target interactions (DTIs) prediction plays an important role in the process of drug discovery. Most computational methods treat it as a binary problem, determining whether there are connections between drugs and targets while ignoring relational types information. Considering positive or negative effects DTIs will facilitate study on comprehensive mechanisms multiple common target, this work, we model signed heterogeneous networks, through categorizing interaction patterns additionally extracting within pairs target protein pairs. We propose graph neural networks (SHGNNs), further put forward end-to-end framework for prediction, called SHGNN-DTI, which not only adapts to bipartite but also could naturally incorporate auxiliary information from drug-drug (DDIs) protein-protein (PPIs). For framework, solve message passing aggregation problem DTI consider different training modes whole consisting DTIs, DDIs PPIs. Experiments conducted two datasets extracted DrugBank related databases, under settings initial inputs, embedding dimensions modes. The results show excellent performance terms metric indicators, feasibility is verified by case with breast cancer.

Language: Английский

Citations

9
Knowledge Graph Convolutional Network with Heuristic Search for Drug Repositioning DOI

Xiang Du,

Xinliang Sun, Min Li

et al.

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: 64(12), P. 4928 - 4937

Published: June 5, 2024

Drug repositioning is a strategy of repurposing approved drugs for treating new indications, which can accelerate the drug discovery process, reduce development costs, and lower safety risk. The advancement biotechnology has significantly accelerated speed scale biological data generation, offering significant potential through biomedical knowledge graphs that integrate diverse entities relations from various sources. To fully learn semantic information topological structure graph, we propose graph convolutional network with heuristic search, named KGCNH, effectively utilize diversity relationships in graphs, as well information, to predict associations between diseases. Specifically, design relation-aware attention mechanism compute scores each neighboring entity given under different relations. address challenge randomness initial potentially impacting model performance expand search scope model, designed module based on Gumbel-Softmax, uses introduces assist exploring more optimal embeddings Following this module, derive relation weights, obtain diseases neighborhood aggregation, then drug–disease associations. Additionally, employ feature-based augmented views enhance robustness mitigate overfitting issues. We have implemented our method conducted experiments two sets. results demonstrate KGCNH outperforms competing methods. In particular, case studies lithium quetiapine confirm retrieve actual top prediction results.

Language: Английский

Citations

5
TarIKGC: A Target Identification Tool Using Semantics-Enhanced Knowledge Graph Completion with Application to CDK2 Inhibitor Discovery DOI

Shen Xiao-juan,

Shijia Yan,

Tao Zeng

et al.

Journal of Medicinal Chemistry, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 2, 2025

Target identification is a critical stage in the drug discovery pipeline. Various computational methodologies have been dedicated to enhancing classification performance of compound-target interactions, yet significant room remains for improving recommendation performance. To address this challenge, we developed TarIKGC, tool target prioritization that leverages semantics enhanced knowledge graph (KG) completion. This method harnesses representation learning within heterogeneous compound-target-disease network. Specifically, TarIKGC combines an attention-based aggregation neural network with multimodal feature extractor simultaneously learn internal semantic features from biomedical entities and topological KG. Furthermore, KG embedding model employed identify missing relationships among compounds targets. In silico evaluations highlighted superior repositioning tasks. addition, successfully identified two potential cyclin-dependent kinase 2 (CDK2) inhibitors novel scaffolds through reverse fishing. Both exhibited antiproliferative activities across multiple therapeutic indications targeting CDK2.

Language: Английский

Citations

0
Integrating machine learning and deep learning algorithms in knowledge graph for disease screening and cataloging: Tools and approaches for drug invention and additive manufacturing DOI
Bhupinder Singh, Christian Kaunert

Elsevier eBooks, Journal Year: 2025, Volume and Issue: unknown, P. 181 - 193

Published: Jan. 1, 2025

Language: Английский

Citations

0
Quality-Controllable automatic construction method of Chinese knowledge graph for medical decision-making applications DOI
Xue Li, Ye Yuan, Yang Yang

et al.

Information Processing & Management, Journal Year: 2025, Volume and Issue: 62(4), P. 104148 - 104148

Published: March 23, 2025

Language: Английский

Citations

0
An Automatic Generation of Heterogeneous Knowledge Graph for Global Disease Support: A Demonstration of a Cancer Use Case DOI Creative Commons

Noura Maghawry,

Samy Ghoniemy, Eman Shaaban

et al.

Big Data and Cognitive Computing, Journal Year: 2023, Volume and Issue: 7(1), P. 21 - 21

Published: Jan. 24, 2023

Semantic data integration provides the ability to interrelate and analyze information from multiple heterogeneous resources. With growing complexity of medical ontologies big generated different resources, there is a need for integrating finding relationships between distinct concepts where these have logical relationships. Standardized Medical Ontologies are explicit specifications shared conceptualization, which provide predefined vocabulary that serves as stable conceptual interface sources. Intelligent Healthcare systems such disease prediction require reliable knowledge base based on ontologies. Knowledge graphs emerged powerful dynamic representation base. In this paper, framework proposed automatic graph generation two standardized ontologies- Human Disease Ontology (DO), Symptom (SYMP) using online website encyclopedia. The methodologies adopted automatically generating fully integrated dynamic, scalable, easily reproducible, reliable, practically efficient. A subgraph cancer terms also extracted studied modeling representing diseases, their symptoms, prevention, risk factors.

Language: Английский

Citations

10
Few-shot biomedical relation extraction using data augmentation and domain information DOI
Bicheng Guo, Di Zhao, Xin Dong

et al.

Neurocomputing, Journal Year: 2024, Volume and Issue: 595, P. 127881 - 127881

Published: May 17, 2024

Language: Английский

Citations

3