Using language models and ontology topology to perform semantic mapping of traits between biomedical datasets

Yi Liu; Benjamin Elsworth; Tom R Gaunt

doi:10.1093/bioinformatics/btad169

Using language models and ontology topology to perform semantic mapping of traits between biomedical datasets

Yi Liu, Benjamin Elsworth, Tom R Gaunt^*

^*Corresponding author for this work

Research output: Contribution to journal › Article (Academic Journal) › peer-review

3 Citations (Scopus)

48 Downloads (Pure)

Abstract

Motivation Human traits are typically represented in both the biomedical literature and large population studies as descriptive text strings. Whilst a number of ontologies exist, none of these perfectly represent the entire human phenome and exposome. Mapping trait names across large datasets is therefore time-consuming and challenging. Recent developments in language modelling have created new methods for semantic representation of words and phrases, and these methods offer new opportunities to map human trait names in the form of words and short phrases, both to ontologies and to each other. Here we present a comparison between a range of established and more recent language modelling approaches for the task of mapping trait names from UK Biobank to the Experimental Factor Ontology (EFO), and also explore how they compare to each other in direct trait-to-trait mapping.

Results In our analyses of 1191 traits from UK Biobank with manual EFO mappings, the BioSentVec model performed best at predicting these, matching 40.3% of the manual mappings correctly. The BlueBERT-EFO model (finetuned on EFO) performed nearly as well (38.8% of traits matching the manual mapping). In contrast, Levenshtein edit distance only mapped 22% of traits correctly. Pairwise mapping of traits to each other demonstrated that many of the models can accurately group similar traits based on their semantic similarity.

Availability and Implementation Our code is available at https://github.com/MRCIEU/vectology.

Original language	English
Article number	btad169
Number of pages	7
Journal	Bioinformatics
Volume	39
Issue number	4
DOIs	https://doi.org/10.1093/bioinformatics/btad169
Publication status	Published - 3 Apr 2023

Bibliographical note

Funding Information:
This work was supported by the UK Medical Research Council Integrative Epidemiology Unit at the University of Bristol [MC_UU_00011/4]. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising.

Publisher Copyright:
© The Author(s) 2023. Published by Oxford University Press.

Keywords

data integration
information retrieval
knowledge representation
machine learning
natural language processing
ontology

Access to Document

10.1093/bioinformatics/btad169

btad169Final published version, 1.05 MBLicence: CC BY

Handle.net

Persistent link

Cite this

@article{0f47303296a94bd984b5c2ba9733bb22,

title = "Using language models and ontology topology to perform semantic mapping of traits between biomedical datasets",

abstract = "Motivation Human traits are typically represented in both the biomedical literature and large population studies as descriptive text strings. Whilst a number of ontologies exist, none of these perfectly represent the entire human phenome and exposome. Mapping trait names across large datasets is therefore time-consuming and challenging. Recent developments in language modelling have created new methods for semantic representation of words and phrases, and these methods offer new opportunities to map human trait names in the form of words and short phrases, both to ontologies and to each other. Here we present a comparison between a range of established and more recent language modelling approaches for the task of mapping trait names from UK Biobank to the Experimental Factor Ontology (EFO), and also explore how they compare to each other in direct trait-to-trait mapping.Results In our analyses of 1191 traits from UK Biobank with manual EFO mappings, the BioSentVec model performed best at predicting these, matching 40.3% of the manual mappings correctly. The BlueBERT-EFO model (finetuned on EFO) performed nearly as well (38.8% of traits matching the manual mapping). In contrast, Levenshtein edit distance only mapped 22% of traits correctly. Pairwise mapping of traits to each other demonstrated that many of the models can accurately group similar traits based on their semantic similarity.Availability and Implementation Our code is available at https://github.com/MRCIEU/vectology.",

keywords = "data integration, information retrieval, knowledge representation, machine learning, natural language processing, ontology",

author = "Yi Liu and Benjamin Elsworth and Gaunt, {Tom R}",

note = "Funding Information: This work was supported by the UK Medical Research Council Integrative Epidemiology Unit at the University of Bristol [MC_UU_00011/4]. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising. Publisher Copyright: {\textcopyright} The Author(s) 2023. Published by Oxford University Press.",

year = "2023",

month = apr,

day = "3",

doi = "10.1093/bioinformatics/btad169",

language = "English",

volume = "39",

journal = "Bioinformatics",

issn = "1367-4811",

publisher = "Oxford University Press",

number = "4",

}

TY - JOUR

T1 - Using language models and ontology topology to perform semantic mapping of traits between biomedical datasets

AU - Liu, Yi

AU - Elsworth, Benjamin

AU - Gaunt, Tom R

N1 - Funding Information: This work was supported by the UK Medical Research Council Integrative Epidemiology Unit at the University of Bristol [MC_UU_00011/4]. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising. Publisher Copyright: © The Author(s) 2023. Published by Oxford University Press.

PY - 2023/4/3

Y1 - 2023/4/3

N2 - Motivation Human traits are typically represented in both the biomedical literature and large population studies as descriptive text strings. Whilst a number of ontologies exist, none of these perfectly represent the entire human phenome and exposome. Mapping trait names across large datasets is therefore time-consuming and challenging. Recent developments in language modelling have created new methods for semantic representation of words and phrases, and these methods offer new opportunities to map human trait names in the form of words and short phrases, both to ontologies and to each other. Here we present a comparison between a range of established and more recent language modelling approaches for the task of mapping trait names from UK Biobank to the Experimental Factor Ontology (EFO), and also explore how they compare to each other in direct trait-to-trait mapping.Results In our analyses of 1191 traits from UK Biobank with manual EFO mappings, the BioSentVec model performed best at predicting these, matching 40.3% of the manual mappings correctly. The BlueBERT-EFO model (finetuned on EFO) performed nearly as well (38.8% of traits matching the manual mapping). In contrast, Levenshtein edit distance only mapped 22% of traits correctly. Pairwise mapping of traits to each other demonstrated that many of the models can accurately group similar traits based on their semantic similarity.Availability and Implementation Our code is available at https://github.com/MRCIEU/vectology.

AB - Motivation Human traits are typically represented in both the biomedical literature and large population studies as descriptive text strings. Whilst a number of ontologies exist, none of these perfectly represent the entire human phenome and exposome. Mapping trait names across large datasets is therefore time-consuming and challenging. Recent developments in language modelling have created new methods for semantic representation of words and phrases, and these methods offer new opportunities to map human trait names in the form of words and short phrases, both to ontologies and to each other. Here we present a comparison between a range of established and more recent language modelling approaches for the task of mapping trait names from UK Biobank to the Experimental Factor Ontology (EFO), and also explore how they compare to each other in direct trait-to-trait mapping.Results In our analyses of 1191 traits from UK Biobank with manual EFO mappings, the BioSentVec model performed best at predicting these, matching 40.3% of the manual mappings correctly. The BlueBERT-EFO model (finetuned on EFO) performed nearly as well (38.8% of traits matching the manual mapping). In contrast, Levenshtein edit distance only mapped 22% of traits correctly. Pairwise mapping of traits to each other demonstrated that many of the models can accurately group similar traits based on their semantic similarity.Availability and Implementation Our code is available at https://github.com/MRCIEU/vectology.

KW - data integration

KW - information retrieval

KW - knowledge representation

KW - machine learning

KW - natural language processing

KW - ontology

U2 - 10.1093/bioinformatics/btad169

DO - 10.1093/bioinformatics/btad169

M3 - Article (Academic Journal)

C2 - 37010521

SN - 1367-4811

VL - 39

JO - Bioinformatics

JF - Bioinformatics

IS - 4

M1 - btad169

ER -

Using language models and ontology topology to perform semantic mapping of traits between biomedical datasets

Abstract

Bibliographical note

Keywords

Access to Document

Handle.net

Fingerprint

Cite this