Some MeSH resources

A simple library that makes available the Medical Subject Headings (MeSH) vocabulary and tree structure in simple data frame format. Tables for both Descriptor Records and Supplementary Concept Records are included. Pharmacological Actions have have been extracted from both descriptor and supplementary concept files, and collated in a single table. Lastly, descriptor-level word embeddings derived in/made available by Noh & Kavuluru (2021) are included. The R code for XML extraction and restructuring processes is available here. These data are utilitzed in the R package pubmedtk.

MeSH ontology

Based on MeSH files desc2024, mtrees2024 & supp2024.

descriptor-terms

Raw data

readRDS('data/data_mesh_thesaurus.rds') |>
  head() |> knitr::kable()

DescriptorUI	DescriptorName	ConceptUI	TermUI	TermName	ConceptPreferredTermYN	IsPermutedTermYN	LexicalTag	RecordPreferredTermYN
D000001	Calcimycin	M0000001	T000002	Calcimycin	Y	N	NON	Y
D000001	Calcimycin	M0000001	T001124965	4-Benzoxazolecarboxylic acid, 5-(methylamino)-2-((3,9,11-trimethyl-8-(1-methyl-2-oxo-2-(1H-pyrrol-2-yl)ethyl)-1,7-dioxaspiro(5.5)undec-2-yl)methyl)-, (6S-(6alpha(2S*,3S*),8beta(R*),9beta,11alpha))-	N	N	NON	N
D000001	Calcimycin	M0353609	T000001	A-23187	Y	N	LAB	N
D000001	Calcimycin	M0353609	T000001	A 23187	N	Y	LAB	N
D000001	Calcimycin	M0353609	T000004	A23187	N	N	LAB	N
D000001	Calcimycin	M0353609	T000003	Antibiotic A23187	N	N	NON	N

descriptor-tree-numbers

Raw data

readRDS('data/data_mesh_trees.rds') |>
  head() |> knitr::kable()

DescriptorUI	DescriptorName	tree_location	code	cats	mesh1	mesh2	tree1	tree2
D000001	Calcimycin	D02.355.291.933.125	D	Chemicals and Drugs	Organic Chemicals	Ethers	D02	D02.355
D000001	Calcimycin	D02.540.576.625.125	D	Chemicals and Drugs	Organic Chemicals	Lactones	D02	D02.540
D000001	Calcimycin	D03.633.100.221.173	D	Chemicals and Drugs	Heterocyclic Compounds	Heterocyclic Compounds, Fused-Ring	D03	D03.633
D000001	Calcimycin	D04.345.241.654.125	D	Chemicals and Drugs	Polycyclic Compounds	Macrocyclic Compounds	D04	D04.345
D000001	Calcimycin	D04.345.674.625.125	D	Chemicals and Drugs	Polycyclic Compounds	Macrocyclic Compounds	D04	D04.345
D000002	Temefos	D02.705.400.625.800	D	Chemicals and Drugs	Organic Chemicals	Organophosphorus Compounds	D02	D02.705

supplemental-terms

Raw data

readRDS('data/data_scr_thesaurus.rds') |>
  head() |> knitr::kable()

DescriptorUI	DescriptorName	ConceptUI	TermUI	TermName	ConceptPreferredTermYN	IsPermutedTermYN	LexicalTag	RecordPreferredTermYN
C114158	quantum dye macrocyclic europium-chelate	M0294520	T324525	quantum dye macrocyclic europium-chelate	Y	N	NON	Y
C114158	quantum dye macrocyclic europium-chelate	M0294520	T324524	QD macrocyclic	N	N	NON	N
C008718	osteoclast activating factor	M0052808	T082811	osteoclast activating factor	Y	N	NON	Y
C008720	otoline	M0052812	T082815	otoline	Y	N	NON	Y
C002540	miracil A	M0043189	T073192	miracil A	Y	N	NON	Y
C055240	Leakadine	M0155620	T185625	Leakadine	Y	N	TRD	Y

Pharmacological Actions

For drugs included in both MeSH-proper and Supplementary Concept Records.

readRDS('data/data_pharm_action.rds') |>
  head() |> knitr::kable()

DescriptorUI	DescriptorName	PharmActionUI	PharmActionName
D000001	Calcimycin	D000900	Anti-Bacterial Agents
D000001	Calcimycin	D061207	Calcium Ionophores
D000002	Temefos	D007306	Insecticides
D000040	Abscisic Acid	D010937	Plant Growth Regulators
D000068180	Aripiprazole	D000928	Antidepressive Agents
D000068180	Aripiprazole	D014150	Antipsychotic Agents

Notes & useful links:

• MeSH XML data elements

• XML elements as RDF

• Utility functions

• Supplementary Concept Records

‘SCR records are created for some chemicals, drugs, and other concepts such as rare diseases. They are labeled as “MeSH Supplementary Concept Data” and the unique ID begins with the letter “C.”’

‘Supplementary Concept Records - these are not full MeSH Headings and do not fall under the MeSH tree hierarchy. Many times they are used to identify substances that are not included in the MeSH terms.’

‘These do not belong to the controlled vocabulary as such and are not used for indexing MEDLINE articles; instead they enlarge the thesaurus and contain links to the closest fitting descriptor to be used in a MEDLINE search. Many of these records describe chemical substances.’

Descriptor embeddings

Noh, J., & Kavuluru, R. (2021). Improved biomedical word embeddings in the transformer era. Journal of Biomedical Informatics, 120, 103867.

https://zenodo.org/record/4383195#.Y1wDBb7MJhE

“BERT-CRel is a transformer model for fine-tuning biomedical word embeddings that are jointly learned along with concept embeddings using a pre-training phase with fastText and a fine-tuning phase with a transformer setup. The goal is to provide high quality pre-trained biomedical embeddings that can be used in any downstream task by the research community. The corpus used for BERT-CRel contains biomedical citations from PubMed and the concepts are from the Medical Subject Headings (MeSH codes) terminology used to index citations.”

readRDS('data/data_scr_embeddings.rds')
readRDS('data/data_mesh_embeddings.rds')