Dataset Information

Linking chemical and disease entities to ontologies by integrating PageRank with extracted relations from literature.

ABSTRACT:

Background

Named Entity Linking systems are a powerful aid to the manual curation of digital libraries, which is getting increasingly costly and inefficient due to the information overload. Models based on the Personalized PageRank (PPR) algorithm are one of the state-of-the-art approaches, but these have low performance when the disambiguation graphs are sparse.

Findings

This work proposes a Named Entity Linking framework designated by Relation Extraction for Entity Linking (REEL) that uses automatically extracted relations to overcome this limitation. Our method builds a disambiguation graph, where the nodes are the ontology candidates for the entities and the edges are added according to the relations established in the text, which the method extracts automatically. The PPR algorithm and the information content of each ontology are then applied to choose the candidate for each entity that maximises the coherence of the disambiguation graph. We evaluated the method on three gold standards: the subset of the CRAFT corpus with ChEBI annotations (CRAFT-ChEBI), the subset of the BC5CDR corpus with disease annotations from the MEDIC vocabulary (BC5CDR-Diseases) and the subset with chemical annotations from the CTD-Chemical vocabulary (BC5CDR-Chemicals). The F1-Score achieved by REEL was 85.8%, 80.9% and 90.3% in these gold standards, respectively, outperforming baseline approaches.

Conclusions

We demonstrated that RE tools can improve Named Entity Linking by capturing semantic information expressed in text missing in Knowledge Bases and use it to improve the disambiguation graph of Named Entity Linking models. REEL can be adapted to any text mining pipeline and potentially to any domain, as long as there is an ontology or other knowledge Base available.

SUBMITTER: Ruas P

PROVIDER: S-EPMC7507273 | biostudies-literature | 2020 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Linking chemical and disease entities to ontologies by integrating PageRank with extracted relations from literature.

Ruas Pedro P Lamurias Andre A Couto Francisco M FM

Journal of cheminformatics 20200921 1

<h4>Background</h4>Named Entity Linking systems are a powerful aid to the manual curation of digital libraries, which is getting increasingly costly and inefficient due to the information overload. Models based on the Personalized PageRank (PPR) algorithm are one of the state-of-the-art approaches, but these have low performance when the disambiguation graphs are sparse.<h4>Findings</h4>This work proposes a Named Entity Linking framework designated by Relation Extraction for Entity Linking (REEL ...[more]

PMID: 33430995

Similar Datasets

Project description:BackgroundThe Centre for Therapeutic Target Validation (CTTV - https://www.targetvalidation.org/) was established to generate therapeutic target evidence from genome-scale experiments and analyses. CTTV aims to support the validity of therapeutic targets by integrating existing and newly-generated data. Data integration has been achieved in some resources by mapping metadata such as disease and phenotypes to the Experimental Factor Ontology (EFO). Additionally, the relationship between ontology descriptions of rare and common diseases and their phenotypes can offer insights into shared biological mechanisms and potential drug targets. Ontologies are not ideal for representing the sometimes associated type relationship required. This work addresses two challenges; annotation of diverse big data, and representation of complex, sometimes associated relationships between concepts.MethodsSemantic mapping uses a combination of custom scripting, our annotation tool 'Zooma', and expert curation. Disease-phenotype associations were generated using literature mining on Europe PubMed Central abstracts, which were manually verified by experts for validity. Representation of the disease-phenotype association was achieved by the Ontology of Biomedical AssociatioN (OBAN), a generic association representation model. OBAN represents associations between a subject and object i.e., disease and its associated phenotypes and the source of evidence for that association. The indirect disease-to-disease associations are exposed through shared phenotypes. This was applied to the use case of linking rare to common diseases at the CTTV.ResultsEFO yields an average of over 80% of mapping coverage in all data sources. A 42% precision is obtained from the manual verification of the text-mined disease-phenotype associations. This results in 1452 and 2810 disease-phenotype pairs for IBD and autoimmune disease and contributes towards 11,338 rare diseases associations (merged with existing published work [Am J Hum Genet 97:111-24, 2015]). An OBAN result file is downloadable at http://sourceforge.net/p/efo/code/HEAD/tree/trunk/src/efoassociations/. Twenty common diseases are linked to 85 rare diseases by shared phenotypes. A generalizable OBAN model for association representation is presented in this study.ConclusionsHere we present solutions to large-scale annotation-ontology mapping in the CTTV knowledge base, a process for disease-phenotype mining, and propose a generic association model, 'OBAN', as a means to integrate disease using shared phenotypes.AvailabilityEFO is released monthly and available for download at http://www.ebi.ac.uk/efo/.

Project description:BackgroundThe Basic Formal Ontology (BFO) is a top-level formal foundational ontology for the biomedical domain. It has been developed with the purpose to serve as an ontologically consistent template for top-level categories of application oriented and domain reference ontologies within the Open Biological and Biomedical Ontologies Foundry (OBO). BFO is important for enabling OBO ontologies to facilitate in reliably communicating and managing data and metadata within and across biomedical databases. Following its intended single inheritance policy, BFO's three top-level categories of material entity (i.e. 'object', 'fiat object part', 'object aggregate') must be exhaustive and mutually disjoint. We have shown elsewhere that for accommodating all types of constitutively organized material entities, BFO must be extended by additional categories of material entity.Methodology/principal findingsUnfortunately, most biomedical material entities are cumulative-constitutively organized. We show that even the extended BFO does not exhaustively cover cumulative-constitutively organized material entities. We provide examples from biology and everyday life that demonstrate the necessity for 'portion of matter' as another material building block. This implies the necessity for further extending BFO by 'portion of matter' as well as three additional categories that possess portions of matter as aggregate components. These extensions are necessary if the basic assumption that all parts that share the same granularity level exhaustively sum to the whole should also apply to cumulative-constitutively organized material entities. By suggesting a notion of granular representation we provide a way to maintain the single inheritance principle when dealing with cumulative-constitutively organized material entities.Conclusions/significanceWe suggest to extend BFO to incorporate additional categories of material entity and to rearrange its top-level material entity taxonomy. With these additions and the notion of granular representation, BFO would exhaustively cover all top-level types of material entities that application oriented ontologies may use as templates, while still maintaining the single inheritance principle.

Dataset Information

Linking chemical and disease entities to ontologies by integrating PageRank with extracted relations from literature.

Background

Findings

Conclusions

Publications

Linking chemical and disease entities to ontologies by integrating PageRank with extracted relations from literature.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets