MetaboliteNET: a metabolite centric knowledge base by ontology-based mining annotations from literature
Description
Recent estimates in dementia cases projected that the largest increases due to population ageing are observed in east Asia, and north Africa and the Middle East [1]. There is an increasing body of scientific evidence on how natural medicines (i.e. herbal medicines, traditional formulas) and food metabolites are environmental factors that can be used to maintain health or treat multifactorial and chronic diseases such as neurodegenerative diseases [2]. For example, traditional Arabic herbal medicine is a part of modern life in the Middle East, and it is acquiring worldwide respect, with growing interest among traditional herbalists and the scientific community. Traditional medicines represent likely sources of relatively inexpensive drugs for symptomatic management as well as potential libraries of new therapeutic approaches. Plants produce a wide diversity of metabolites that exhibit a wide array of biological and pharmacological properties. The multitarget activities of many metabolites can explain the medical application of complex extracts from medicinal plants, food or even drug metabolites for health disorders that involve several targets. Therefore, the development of a metabolite-centric annotation dataset from literature is useful for the identification of functionally similar metabolites with application in disease biomarker discovery, drug discovery or agriculture. Furthermore, Data science pipelines applied on text mining based knowledge graphs have the potential to accelerate understanding on mechanisms and mode of actions, link therapies to new potential target applications, and rapidly accelerate the validation (and exclusion) of traditional medicines [3].
The motivation of this project is to provide a metabolite-centric annotation dataset as a valuable knowledge resource to enhance precision medicine in the MENA region. We propose the development of a metabolite centric annotation data set, the MetaboliteNET annotation data set, built from the literature and existing resources for machine learning prediction of interesting metabolites.
Our proposal is to extract human and plant metabolite annotations from literature and make them publicly available. We aim at building a knowledge graph from the metabolite - X annotation data set, where X will be phenotypes (Human Phenotype Ontology or HPO), diseases (Human Disease Ontology or DO, and Mondo Disease Ontology or MONDO), environmental factors (Environment Ontology or ENVO), anatomical (Uberon multi-species anatomy Ontology or UBERON), cell types (Cell Ontology or CL), gene products functions (Gene Ontology or GO) for human metabolites, and X will be ENVO, CL, phenotypes (Flora Phenotype Ontology or FLOPO, and Plant Trait Ontology or PTO), anatomy (Plant Ontology or PO) for plants. The MetaboliteNET annotation data set will be yielded automatically from mining the literature using an in-house annotation tool that performs co-occurrence extraction of mentions (as ontology terms) at abstract level over all MEDLINE, and applies a statistical approach based on ontological structure to determine the association strength [4]. This tool is currently being updated to extract relevant information at mention and publication level. We will apply the license Creative Commons Attribution 4.0 International to the annotation data set. We will evaluate our text mined data against curated associations. The MetaboliteNET annotation data set will serve as a foundation to build a metabolite-centric knowledge graph for representation learning or curation for corpora development.
[1] GBD 2019 Dementia Forecasting Collaborators, Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019, The Lancet, 7, 2, e105-e125, February 01, 2022, https://doi.org/10.1016/S2468-2667(21)00249-8 [2] Emmanuel Ayodeji Ayeni, Yuzhou Gong, Hao Yuan, Yikao Hu, Xiaolin Bai, Xun Liao, Medicinal Plants for Anti-neurodegenerative diseases in West Africa, Journal of Ethnopharmacology, 285, 2022, 114468, https://doi.org/10.1016/j.jep.2021.114468. [3] C. Jansen, J.D. Baker, E. Kodaira, L. Ang, A.J. Bacani, J.T. Aldan, L.M.N. Shimoda, M. Salameh, A.L. Small-Howard, A.J. Stokes, H. Turner, C.N. Adra, Medicine in motion: Opportunities, challenges and data analytics-based solutions for traditional medicine integration into western medical practice, Journal of Ethnopharmacology, Volume 267, 2021, https://doi.org/10.1016/j.jep.2020.113477. [4] Şenay Kafkas, Robert Hoehndorf, Ontology based text mining of gene-phenotype associations: application to candidate gene prediction, Database, Volume 2019, 2019, baz019, https://doi.org/10.1093/database/baz019
Acknowledgements
This project was first proposed and developed during the Biomedical Linked Annotation Hackathon 6 on February 4-7 2020 in Tokyo, Japan. It was further developed during the BioHackathon-MENA 2023 on February 7-11 2023 in KAUST, KSA.
We are very grateful to the BLAH6 and BioHackathon-MENA 2023 organizers for the support and funding on this work.