Search

Démonstration; International audience; Recent advances in high-throughput technologies have revolutionized the analysis in the field of the plant sciences. However, there is an urgent need to effectively integrate and assimilate complementary information to understand the biological system in its entirety. We have developed AgroLD, a knowledge graph that exploits Semantic Web technologies to integrate data of interest for the plant science community e.g., rice, wheat, arabidopsis and in this way facilitate the formulation and validation of new scientific hypotheses. AgroLD contains around 900M triples created by annotating and integrating more than 100 datasets coming from 15 data sources. Our objective is to offer a domain specific knowledge platform to answer complex biological and plant sciences questions related to the implication of genes in, for instance, plant disease resistance or adaptative responses to climate change. In this demo, we present some results which currently focused on genomics, genetics and trait associations.

ISSN : 16130073; International audience; DLinker is a system for matching instances of two RDF data sources. Its performance is mainly based on the deep comparison of literals. The main comparison algorithm is based on the search for the longest common subsequence (LCS) present in the literals. The validation of the similarity between two literals is performed by a mathematical formula. This formula computes the confidence percentage of the similarity between the literals and compares it with a threshold provided as input among the expected hyperparameters. To validate the similar instances, our system validates only the instances that have reached the value of the acceptation threshold provided in the list of required hyperparameters. The current version focuses on the processing of strings on the spot without taking into account synonyms to make its decisions. This is DLinker's first participation in the OAEI campaign on two principal tracks (SPIMBENCH and SPATIAL) with 9 challenges. DLinker demonstrated its ability to process different data with good accuracy and in a very short time. Additionally, in the context of the SPATIAL challenge DLinker has outperformed the state of the art finishing first with the shortest time. Overall, DLinker exposes different strengths and weaknesses that are discussed in this work.

National audience

International audience; The Explore Relationships tool aids in exploring relationships between existing entities. Quick search is based on keyword search and aids in understanding the underlying knowledge.

International audience; Recent advances in high-throughput technologies have resulted in tremendous increase in the amount of data in the agronomic domain. There is an urgent need to effectively integrate complementary information to understand the biological system in its entirety. We have devel- oped AgroLD, a knowledge graph that exploits the Semantic Web technology and some of the relevant standard domain ontologies, to integrate information on plant species and in this way facilitating the formulation of new scientific hypotheses. We present some integration results of the project, which initially focused on genomics, proteomics and phenomics.; Les récents progrès des technologies à haut débit ont entraîné une explosion de la quantité de données dans le domaine agronomique. Il est urgent d'intégrer efficacement des informations complémentaires pour comprendre le système biologique dans sa globalité. Nous avons développé AgroLD, une base de connaissances qui exploite la technologie du Web sémantique et des ontologies du domaine biologique pertinentes, pour intégrer les informations sur les espèces végétales et faciliter ainsi la formulation de nouvelles hypothèses scientifiques. Nous présentons des résultats sur le processus d'intégration et sur la plateforme visualisation des données, qui était initialement axé sur la génomique, la protéomique et la phénomique.

Version deposé et non-validée par les rapporteurs

Summary Currently, gene information available for Oryza sativa species is located in various online heterogeneous data sources. Moreover, methods of access are also diverse, mostly web-based and sometimes query APIs, which might not always be straightforward for domain experts. The challenge is to collect information quickly from these applications and combine it logically, to facilitate scientific research. We developed a Python package named PyRice, a unified programming API to access all supported databases at the same time with consistent output. PyRice design is modular and implements a smart query system which fits the computing resources to optimize the query speed. As a result, PyRice is easy to use and produces intuitive results. Availability and implementation https://github.com/SouthGreenPlatform/PyRice Documentation https://pyrice.readthedocs.io Contact pierre.larmande@ird.fr Licence information MIT Supplementary information Supplementary data are available at Bioinformatics online.

Version deposé et non-validée par les rapporteurs

La compréhension des relations génotype-phénotype est un des axes les plus important de la recherche en agronomie. Or les interactions génotype-phénotype sont complexes à identifier car elles s'expriment à différentes échelles moléculaires dans la plante et subissent de fortes influences de la part des facteurs environnementaux. Les technologies d'analyses haut-débit ne permettent de capturer que partiellement cette dynamique. Même si ces technologies permettent d'aller toujours plus loin dans l'obtention de nouvelles données, notre connaissance reste encore parcelLaire pour élucider les mécanismes moléculaires qui régissent l'expression des caractères phénotypiques complexes. Les nouveaux défis consistent à comprendre les relations complexes existant entre les différents éléments moléculaires responsables de l'expression du phénome. Cet objectif ne peut être atteint qu'en intégrant des informations de différents niveaux dans un modèle intégrateur utilisant une approche systémique afin de comprendre le fonctionnement réel d'un système biologique. Mon projet de recherche aborde le problème suivant : Comment structurer et gérer la complexité des données biologiques afin d'en extraire de la connaissance permettant d'identifier les mécanismes moléculaires contrôlant l'expression de phénotypes chez les plantes. L'objectif de ce projet sera de déterminer si la représentation d'information sous forme de graphes de connaissances est adaptée pour formuler des hypothèses de recherche permettant de lier le génotype au phénotype. En prenant le riz comme modèle, l'objectif sera de construire des réseaux d'interaction moléculaires à partir de données éparses afin d'identifier les gènes clés pour l'amélioration des plantes. Plusieurs approches de recherche sont envisagées : intégration des données, enrichissement des connaissances, applications sur les graphes de connaissances. Dans ce processus, une première voie consistera à transformer et intégrer dynamiquement ces données dans la base de connaissance AgroLD pour les rendre plus facilement utilisables en terme algorithmique. Une deuxième voie consistera à proposer de nouvelles méthodes d'enrichissement des connaissances. Dans un premier temps, en se focalisant sur des méthodes d'annotation sémantique. Puis, afin d'enrichir les liens entre les différents graphes générés et ainsi produire un réseau d'interaction qui permettra la découverte de nouvelles connaissances, de nouvelles méthodes de liage de données seront développées. Enfin, afin de permettre une recherche d'information efficace, plusieurs méthodes et algorithmes de priorisation de gènes candidats seront évaluées et proposées.

Agronomy/agriculture and biodiversity (ag & biodiv) communities face several major societal, economic, and environmental challenges that data science approaches will help address. To achieve their goals, researchers of these communities must be able to rapidly discover, aggregate, integrate, and analyse different types of data and information sources. Semantic technologies, combined to open, FAIR data and services, is one of the answers to fully knowledge-driven, and transparent science and innovation. The D2KAB project (www.d2kab.org) aims to create a framework to turn agronomy and biodiversity data into knowledge – semantically described, interoperable, actionable, open – and investigate the scientific methods and tools to exploit this knowledge for applications in agriculture and biodiversity sciences. This project, funded by French ANR (2019-2023), will provide the means –ontologies and linked open data– for ag & biodiv to embrace semantic Web technologies in order to produce and exploit FAIR data and services. To do so, D2KAB will develop new original methods and algorithms in the following areas: data integration, text mining, semantic annotation, ontology alignment and linked data exploitation and visualization. D2KAB project brings together a unique multidisciplinary consortium of 12 partners to achieve this objective: 2 informatics research units (LIRMM, I3S); 6 INRA/IRSTEA/IRD research units at the interface of computer science and ag & biodiv (URGI, MaIAGE, IATE, DIST, TSCF, DIADE) specialized in agronomy or agriculture; 2 labs in biodiversity and ecosystem research (CEFE, URFM); 1 association of agriculture stakeholders (ACTA); and 1 partnership with Stanford BMIR department. Three main goals drive D2KAB’s roadmap: To develop state-of-the-art methods and technologies for ontology lifecycle and alignment. To build the agronomy, agriculture and biodiversity Linked Open Data cloud. To enable new semantically driven agronomy and biodiversity science. The work is starting from the recommendations of several RDA WG and IG already published or in progress (e.g. Agrisemantic WG, Vocabulary Services IG, Wheat and Rice Data Interoperability WGs, Agricultural Data IG, SHARC IG). Some of the key technological building blocks of D2KAB are AgroPortal, a reference repository for ontologies and vocabularies in agronomy; AgroLD, a semantic Web knowledge base that integrates agronomic data from public databases including GO associations, Gramene, UniprotKB, and OryGenesDB ; Corese, a semantic Web factory that implements the W3C standards RDF, RDFS, OWL-RL and SPARQL, and LDScript, a Linked Data Script Language, and STTL, the SPARQL Template Transformation Language for RDF; and Alvis, a text mining for semantic normalisation of free text by ontologies. D2KAB will allow the valorization of ag & biodiv data into real world applications leading to economic impact, smart agriculture and ecological preservation. Five driving scenarios are planned: development of an ontology-based expert system to select food packaging solutions; creation of an augmented semantic reader for Plant Health Bulletins; advanced integration of textual and experimental data on wheat phenotypes; development of new ontologies on plant root traits and extension of the Thesaurus Of Plant Characteristics; integration of plant functional biogeography data related to the Mediterranean Basin. Each of the project scenarios will have a significant impact and produce concrete outcomes for ag & biodiv scientific communities and socio-economic stakeholders in agriculture.

National audience; La compréhension des interactions génotype-phénotype est un des axes les plus importants de la recherche en agronomie dont l'un des objectifs est d'accélérer la reproduction des caractères importants pour la production agricole. Or ces interactions sont complexes à identifier car elles s'expriment à différentes échelles moléculaires dans la plante et subissent de fortes influences de la part des facteurs environnemen-taux. Les technologies d'analyse haut-débit ne permettent de capturer que partiellement cette dynamique. Même si ces technologies sont de plus en plus performantes dans l'acquisition de données, notre connais-sance du système reste encore parcellaire pour pouvoir comprendre les relations complexes existant entre les différents éléments moléculaires responsables de l'expression du phénome-ensemble des phénotypes observés pour un individu-. Cet objectif ne peut être atteint qu'en intégrant des informations de différents niveaux dans un modèle intégrateur utilisant une approche systémique afin de comprendre le fonctionnement réel d'un système biologique. Aujourd'hui, le Web sémantique propose des technologies pour l'intégration de données hétérogènes et leur transformation en connaissances explicites grâce aux ontologies. Nous avons développé AgroLD (Venkatesan et al., 2018) (Agronomic Linked Data-www.agrold.org), une base de connaissances reposant sur les technologies du Web sémantique et exploitant des ontologies du domaine biologique, afin d'intégrer des données issues de plusieurs espèces de plantes présentant un intérêt important pour la communauté scientifique, comme par exemple le riz, le blé et arabidopsis. Nous présentons les résultats du projet, qui portait initialement sur la génomique, la protéomique et la phénomique. AgroLD est aujourd'hui une base de plus de 100 millions de triplets créée à partir de plus de 50 jeux de données provenant d'une dizaine de sources de données, telles que Gramene (Tello-Ruiz et al., 2018) et TropGeneDB (Hamelin et al., 2012). Par ailleurs, nous avons utilisé une dizaine d'ontologies du domaine biologique, telles que Gene Ontology (The Gene Ontology Consortium, 2014) et Plant Ontology (Plant & Consortium, 2002) pour annoter et intégrer ces ressources. Pour cette phase, chaque jeu de données a été transformé à partir de sources sélectionnées et annotées sémantiquement en réutilisant les champs textuels correspondant avec des termes d'ontologies lorsqu'ils ont été fournis par la source d'origine. De plus, nous avons utilisé les services Web d'AgroPortal (Jonquet et al., 2018) pour annoter sémantiquement des éléments supplémentaires tels que par exemple, les URIs correspondant à la taxonomie des espèces ou des éléments d'anatomie. Dans ces cas, nous avons généré des propriétés supplémentaires à partir des ontologies correspondantes, ajoutant ainsi 22% de triplets supplémentaires qui ont été validés manuellement. L'objectif d'AgroLD est d'offrir une plate-forme de connaissances spécifiques du domaine agronomique afin de répondre à des questions biologiques complexes. De telles questions peuvent concerner le rôle de gènes spécifiques dans les mécanismes de résistance aux maladies des plantes ou de caractères de production identifiés à partir des analyses GWAS. Afin de rendre AgroLD accessible par un plus grand nombre d'uti-lisateurs, nous avons également développé une application Web proposant plusieurs interfaces de requêtes. Tout d'abord une interface simple qui permet aux utilisateurs de rechercher par mots-clés sur l'ensemble des valeurs de la base et ainsi de parcourir le contenu d'AgroLD. Puis une interface de recherche avancée qui permet de combiner du texte libre et des filtres à facettes ainsi que des services Web externes proposant ainsi une interface d'agrégation de données distribuées. AgroLD possède également une interface de visualisation des graphes qu'il est possible de configurer pour mettre en valeur certains types de relations. Finalement, un éditeur SPARQL propose un environnement interactif pour formuler des requêtes et manipuler des ré-sultats. Actuellement, de nouveaux jeux de données sont en cours d'intégration. Ils portent sur les réseaux d'interaction protéine-protéine, les facteurs de transcription et réseaux de co-expression afin d'étendre les connaissances sur les mécanismes moléculaires. De nombreux développements sont également réalisés au niveau des interfaces de requêtes, notamment au niveau de la visualisation des graphes afin de fournir des outils plus dynamiques, interactifs et contextualisés. Enfin, une attention particulière est portée sur la qualité des données intégrées. Des méthodes de liage et de machine learning sont développées pour rechercher des liens et des ressources similaires dans la base de connaissances ou dans des ressources externes.

International audience; Background: The study of genetic variations is the basis of many research domains in biology. From genome structure to population dynamics, many applications involve the use of genetic variants. The advent of next-generation sequencing technologies led to such a flood of data that the daily work of scientists is often more focused on data management than data analysis. This mass of genotyping data poses several computational challenges in terms of storage, search, sharing, analysis, and visualization. While existing tools try to solve these challenges, few of them offer a comprehensive and scalable solution. Results: Gigwa v2 is an easy-to-use, species-agnostic web application for managing and exploring high-density genotyping data. It can handle multiple databases and may be installed on a local computer or deployed as an online data portal. It supports various standard import and export formats, provides advanced filtering options, and offers means to visualize density charts or push selected data into various stand-alone or online tools. It implements 2 standard RESTful application programming interfaces, GA4GH, which is health-oriented, and BrAPI, which is breeding-oriented, thus offering wide possibilities of interaction with third-party applications. The project home page provides a list of live instances allowing users to test the system on public data (or reasonably sized user-provided data). Conclusions: This new version of Gigwa provides a more intuitive and more powerful way to explore large amounts of genotyping data by offering a scalable solution to search for genotype patterns, functional annotations, or more complex filtering. Furthermore, its user-friendliness and interoperability make it widely accessible to the life science community.

With the decreasing cost of genome sequencing, many laboratories are increasingly adopting genotyping technologies as a routine component of their analytical workflows, generating large datasets (e.g. VCF files) of genotyping information. Nevertheless, manipulating such large datasets remains a challenge for many scientists. In this context, we developed Gigwa (Genotype Investigator for Genome-Wide Analyses) with the aim of providing a user-friendly system to meet the requirements of scientists who need to filter large datasets and export them into various formats for subsequent analyses. Gigwa is species-agnostic, cross-platform, scalable and easy to deploy. It can be configured to run on a local computer or setup across servers to act as a data portal. It may be used to share data with collaborators while providing means to seek variants of interest based on location, functional annotations or genotype patterns. Based on NoSQL technology, it supports very large datasets (up to tens of millions of genotypes) when configured on suitable hardware. Its most attractive features are: ergonomic interface including user management, numerous import and export formats, powerful filtering engine, interoperability via REST APIs and connection to online or standalone tools. Gigwa now in version 2 (http://gigwa.southgreen.fr), is developed within the scope of South Green bioinformatics, a cross-institute platform and community dedicated to genetics and genomics of tropical and Mediterranean plants, based in Montpellier, France.

Genotype-phenotype interactions play an important role in agronomy. However, these interactions are complex to identify because expressed at different molecular scales and are strongly influenced by environmental factors. Although high-throughput technologies generate more comprehensive insights, our knowledge remains fragmentary to elucidate the molecular mechanisms that govern the expression of complex traits. This objective can be achieved by integrating information from different levels into an integrative model in order to understand the functioning of the whole biological system. Today, the Semantic Web offers technologies for the integration of heterogeneous data and their transformation into explicit knowledge. We have developed AgroLD (www.agrold.org) relying on Semantic Web technologies and exploiting standard domain ontologies, to integrate data about plant species of high interest for the plant science community. AgroLD is now an RDF (Resource Description Format) knowledge base of 130M triples created by annotating and integrating more than 55 datasets with 10 ontologies. AgroLD's objective is to offer a domain-specific knowledge platform to solve complex biological questions related to the implication of genes/proteins in, for instances, plant disease resistance or high yield traits. Thus, the AgroLD platform provides several entry points, in order to allow a large number of users accessing it.

The Genomic Open-Source Breeding informatics initiative (GOBii) has built a genomics data management system that is highly scalable and has focused on data extract performance for large genomics data files. We have benchmarked several SQL and noSQL open-source data management systems with a view to managing large scale genomics data, and have determined that the HDF5 file system outperformed other data management systems both in loading and extract times. In order to also accommodate metadata management, we have designed and developed a hybrid system based on Postgres for sample and marker metadata management and HDF5 for the large genomics files. The HDF5 genomics files are stored in two different orientations to enable rapid extract in either sample or marker-fast formats. The system is flexible enough to be used across different crops and with diverse marker and sequence-based platforms. We are now working to integrate the genomics data extracts with downstream genomic selection applications in Galaxy.

International audience; Recent advances in high-throughput technologies have resulted in a tremendous increase in the amount of omics data produced in plant science. This increase, in conjunction with the heterogeneity and variability of the data, presents a major challenge to adopt an integrative research approach. We are facing an urgent need to effectively integrate and assimilate complementary datasets to understand the biological system as a whole. The Semantic Web offers technologies for the integration of heterogeneous data and their transformation into explicit knowledge thanks to ontologies. We have developed the Agronomic Linked Data (AgroLD-www.agrold.org), a knowledge-based system relying on Semantic Web technologies and exploiting standard domain ontologies, to integrate data about plant species of high interest for the plant science community e.g., rice, wheat, arabidopsis. We present some integration results of the project, which initially focused on genomics, proteomics and phenomics. AgroLD is now an RDF (Resource Description Format) knowledge base of 100M triples created by annotating and integrating more than 50 datasets coming from 10 data sources-such as Gramene.org and TropGeneDB-with 10 ontologies-such as the Gene Ontology and Plant Trait Ontology. Our evaluation results show users appreciate the multiple query modes which support different use cases. AgroLD's objective is to offer a domain specific knowledge platform to solve complex biological and agronomical questions related to the implication of genes/proteins in, for instances, plant disease resistance or high yield traits. We expect the resolution of these questions to facilitate the formulation of new scientific hypotheses to be validated with a knowledge-oriented approach.

International audience; Text mining is becoming an important part of research topic in biology with the original purpose to extract biological entities such as genes, proteins and traits to extend the knowledge from scientific papers. However, few thorough studies on text mining and applications are developed for plant molecular biology data, especially rice, thus resulting a lack of datasets available to train models. Since there is no/rare benchmark for rice, we have to face various difficulties in exploiting advanced machine learning methods for accurate analysis of rice bibliography. In this article, we developed a new training datasets (Oryzabase) as the benchmark. Then, we evaluated the performance of several current approaches to find a methodology with the best results and assigned it as the state of the art method for our own technique in the future. We applied Name Entities Recognition (NER) tagger, which is built from a Long Short Term Memory model, and combined with Conditional Random Fields (LSTM-CRF) to extract information of plant genes and proteins. We analyzed the performance of LSTM-CRF when applying to the Oryzabase dataset and improved the results up to 89% in F1. We found that on average, the result from LSTM-CRF is more exploitable with the new benchmark.

Recent advances in high-throughput technologies have resulted in tremendous increase in the amount of data in the agronomic domain. This data explosion in conjunction with its heterogeneity presents a major challenge in adopting an integrative approach towards research. There is an urgent need to effectively integrate and assimilate complementary information to understand the biological system in its entirety. To this end, the Semantic Web offers a stack of powerful technologies for the integration of information from diverse sources and make knowledge explicit thanks to ontologies. We have developed AgroLD (www.agrold.org), a knowledge based system that exploits the Semantic Web technology and some of the relevant standard domain ontologies, to integrate genome to phenome information on plant species widely studied by the plant science community. We present some integration results of the project, which initially focused on genomics, proteomics and phonemics. Currently, AgroLD contains hundreds millions of triples created by annotating more than 50 datasets coming from major rice databases with some relevant ontologies. Our objective is to offer a domain specific knowledge platform to solve complex biological and agronomical questions related to the implication of genes or proteins in, for instances, plant disease resistance or high yield traits. We expect the resolution of these questions to facilitate the formulation of new scientific hypotheses to be validated with a knowledge-oriented approach.

The International Rice Informatics Consortium (IRIC), established in 2014, is coordinated by the International Rice Research Institute, and currently has 17 private & public sector member institutions. IRIC aims to 1-organize genotype and available phenotype and passport data for rice germplasm into a linked, consistent and reliable source of information for the global research community, 2-support the development of a user-friendly platform to browse, search and analyze rice data through a single Web portal, and 3-coordinate information sharing and provide a communication platform for public awareness and capacity building. To fulfill objectives 1-2, IRIC leads the development of the SNP-Seek database, a web resource that allows rapid searching of genotypes, varieties, and analyses results from the 3,000 Rice Genomes, High-Density Rice Array and various sequencing/ genotyping projects, and public rice genome information. IRIC also contributes to international initiatives for data standards (Crop Ontology, CGIAR Big Data & Genebank Platforms), data/database interoperability (Breeding API, Rice Data Interoperability Working Group), and reproducible analyses (Rice Galaxy, Excellence in Breeding). IRIC conducts outreach, training and collaboration with the rice research community through annual workshops at Plant Animal Genome conference and the International Symposium on Rice Functional Genomics, the IRIC website, and social media through Facebook and LinkedIn. In this talk the various bioinformatics resources of IRIC will be presented in greater detail.

International audience; Plant Phenotyping data management following the FAIR (Findable, Accessible, Interoperable, Resusable) is highly challenging because of its heterogenity. Thus, simply integrating and consolidating data within a single dataset like a phenotyping network is already a complicated task which is even more complex when trying to link different datasets together. To adress this problem, the Minimal Information About Plant Phenotyping Experiment standard construction has been initiated four years ago, with the help of experts from European infrastructures and institutes like Elixir, Emphasis, INRA, WUR, iBet, IPK, EBI and IPG PAS. It adresses the need of data publication and reuse through a checklist that formalize and document the minimal metadata necessary to ensure long term FAIRness of field or greenhouse datasets, including high througputs phenotyping ones. This list has been implemented in several databases like GnpIS or eDale, in a file format, ISA Tab, in a web service, the Breeding API and an RDF implementation is under construction. We will review those implementations, show its current adoption state and detail the plans for the future evolutions of the standard.

Plant Phenotyping data management following the FAIR (Findable, Accessible, Interoperable, Resusable) is highly challenging because of its heterogenity. Thus, simply integrating and consolidating data within a single dataset like a phenotyping network is already a complicated task which is even more complex when trying to link different datasets together. To adress this problem, the Minimal Information About Plant Phenotyping Experiment standard construction has been initiated four years ago, with the help of experts from European infrastructures and institutes like Elixir, Emphasis, INRA, WUR, iBet, IPK, EBI and IPG PAS. It adresses the need of data publication and reuse through a checklist that formalize and document the minimal metadata necessary to ensure long term FAIRness of field or greenhouse datasets, including high througputs phenotyping ones. This list has been implemented in several databases like GnpIS or eDale, in a file format, ISA Tab, in a web service, the Breeding API and an RDF implementation is under construction. We will review those implementations, show its current adoption state and detail the plans for the future evolutions of the standard.