1. Summary of the project
Plants possess an intricate chemo-diversity, serving as a rich reservoir for the discovery of potential therapeutic agents. In the framework of a Swiss research initiative called "Sinergia", six research groups from diferent disciplines collaborate to explore the potential of more than 17,000 unique dried plant extracts, to uncover novel bioactive molecules. To do so, heterogenous data from the diferent research groups was integrated in a Knowledge Graph (KG). This includes (i) highresolution mass spectrometry data, (ii) taxonomical information, (iii) chemo-informatics results, (iv) bioassay outcomes for tuberculosis, obesity, anticancer and antiviral models, as well as (v) data from organic synthetic chemistry.
We wanted to create a comprehensive framework to facilitate the alignment of multimodal data across chemical structures, biological activities, spectral features, and taxonomy, among others. At its core, it would include experimental data that is processed at the sample level, harmonized with external identifiers whenever feasible, semantically enriched and integrated into the KG.
The KG allows to enhance eficient data mining capabilities to address diferent scientific research questions related to the specific objectives of each participating group. For example, "Which bioactive molecules annotated in the extract of ’plant A’ are present in other species of the same genus?", “Which taxonomic group presents a remarkable concentration of bioactive molecules for a specific assay?", or "Which non-toxic active extracts (for a specific assay) present a high probability of containing new molecules?".
The modelling of data from chemical analysis became a subproject in itself: ENPKG [
The heterogeneity, expansion, and evolution of the data throughout time posed a major challenge for the project, centered on managing, integrating, modeling, and efectively sharing data generated by the diferent groups. Also, compliance with a data management plan and adherence to open-source science principles, particularly the FAIR (Findable, Accessible, Interoperable, Reusable) principles, had to be ensured.
To overcome these challenges and handle this intricate data, a custom tool called KGSteward [
For managing the KG, the free version of GraphDB [
Table 1 provides a snapshot of the diferent datasets integrated into the KG, emphasizing their origin. These datasets, presented as RDF graphs, are organized around the central concept of an "analysis run". This documents data executed by the same operator within the same laboratory on a specific date. Should the operator repeat the same assay on another date, it is regarded as a separate analysis run to ensure comprehensive traceability of integrated data.
Provenance information is encoded using the PROV Ontology [
The use of RDF/SPARQL in our project has proven to be highly robust, particularly in handling the dynamic nature of our evolving datasets. These semantic technologies ofer a notable level of flexibility in defining and adapting vocabularies according to our project’s specific requirements.
However, the delicate balance between flexibility and usability was a central concern during the deployment of our KG. The KG structure was modelled from a starting set of data, but due to the quick evolution of the cumulative heterogenous sets it was modified drastically for a better overall fit. The implementation of KGSteward in the evolution process was instrumental, as it bypasses the graphical interface, automatically detects the modified input TTL files, and does selective updates.
This strategic approach reflects the project’s dedication to efectively capture, organize, and adapt to the continuously generated data and concepts. The integration of RDF/SPARQL aligns with our project goals and reflects our commitment to navigate the complexities of contemporary data management challenges in interdisciplinary research endeavors.
The authors are grateful to Green Mission Pierre Fabre, Pierre Fabre Research Institute, Toulouse, France, for establishing and sharing this unique library of extracts. The authors thank the Swiss National Science Foundation for received support for the project (SNF N° CRSII5_189921/1).
S. Zednik, J. Zhao, PROV-O: The PROV Ontology, World Wide Web Consortium, United States, 2013. [6] M. Dumontier, A.J.G. Gray, M.S. Marshall, V. Alexiev, P. Ansell, G. Bader, J. Baran, J.T. Bolleman, A.
Callahan, J. Cruz-Toledo, P. Gaudet, E.A. Gombocz, A.N. Gonzalez-Beltran, P. Groth, M. Haendel, M. Ito, S. Jupp, N. Juty, T. Katayama, N. Kobayashi, K. Krishnaswami, C. Laibe, N. Le Novère, S. Lin, J. Malone, M. Miller, C.J. Mungall, L. Rietveld, S.M. Wimalaratne, A. Yamaguchi, The health care and life sciences community profile for dataset descriptions, PeerJ. 4 (2016) e2331. doi:10.7717/peerj.2331. [7] L.-M. Quiros-Guerrero, L.-F. Nothias, A. Gaudry, L. Marcourt, P.-M. Allard, A. Rutz, B.
David, E.F. Queiroz, J.-L. Wolfender, Inventa: A computational tool to discover structural novelty in natural extracts libraries, Frontiers in Molecular Biosciences. 9 (2022) 1028334. doi:10.3389/fmolb.2022.1028334. [8] P.-M. Allard, A. Gaudry, L.-M. Quirós-Guerrero, A. Rutz, M. Dounoue-Kubo, T.W.N. Walker, E.
Defossez, C. Long, A. Grondin, B. David, J.-L. Wolfender, Open and reusable annotated mass spectrometry dataset of a chemodiverse collection of 1,600 plant extracts, GigaScience. 12 (2022) giac124. doi:10.1093/gigascience/giac124. [9] A. Rutz, M. Sorokina, J. Galgonek, D. Mietchen, E. Willighagen, A. Gaudry, J.G. Graham, R. Stephan, R. Page, J. Vondrášek, C. Steinbeck, G.F. Pauli, J.-L. Wolfender, J. Bisson, P.-M. Allard, The LOTUS initiative for open knowledge management in natural products research, ELife. 11 (2022) e70780. doi:10.7554/eLife.70780.