AGAP, “Genetic Improvement and Adaptation of Mediterranean and Tropical Plants”, a new member unit of #DigitAg

The AGAP joint research unit – Genetic Improvement and Adaptation of Mediterranean and Tropical Plants – joined #DigitAg in September 2018. With its 400 staff members, AGAP is a very large research unit that constitutes an internationally important centre for applied plant biology and genetics. It brings together teams from CIRAD, INRA and Montpellier SupAgro, as well as members of the Institut Français de la Vigne et du Vin (IFV – French Wine and Vine Institute) and the Conservatoire Botanique National Méditerranéen de Porquerolles (CBNMed – National Mediterranean Botanical Conservatory in Porquerolles). An interview with Emmanuel Guiderdoni (CIRAD, Deputy Director of the unit), Evelynes Costes (INRA) and Jean-Marc Lacape (CIRAD), coordinators of thematic fields within the unit.

The biodiversity of cultivated plants presents varied and numerous sources of adaptive capacity. AGAP analyses and enhances the genetic resources of Mediterranean and tropical cultivated plants by studying the structure of their diversity and of their adaptation to abiotic and biotic constraints. The unit produces knowledge in plant biology and genetics for some 20 species and creates varieties adapted to a range of environments and uses, for agriculture in both the North and the South. AGAP also manages a number of technical facilities and platforms (genotyping, phenotyping, bioinformatics, cellular imaging, modelling, greenhouses, etc.) as well as collections of genetic resources from plant species in its BRCs (Biological Resource Centres).

The unit is structured around three thematic fields:

  1. Diversities & Genomes: structure, domestication, environments, societies”, which deals with the analysis of plant diversity maintained within agrosystems and the study of factors of diversification and of diversity maintenance, including societal factors at scales ranging from the genome to the population. This field mobilises scientific expertise in genetics and genomics, but also anthropology and political science, supported by bioinformatics and biomathematics methodologies.
  2. “Development and adaptation of plants and stands”, which studies the genetic and functional bases of the traits involved in tolerance to abiotic and biotic constraints, such as climate factors, disease resistance and product quality. This field mobilises expertise and technologies in genotyping, phenotyping (all observable traits), functional analysis, imaging and modelling.
  3. “Integrative approaches for varietal innovation”, which puts all of the unit’s knowledge to use within plant improvement programmes. Conducted in partnership and with end users, these selection programmes use genetic, genomic and phenotyping tools.

Keywords: biodiversity – improvement – plant selection / genetics / genomics, biology, physiology – ecophysiology of plants / high-throughput phenotyping – imaging – statistics – computer science (plant modelling – image processing – scientific workflows – data source management) – software engineering – interoperability of data

How does AGAP contribute to digital agriculture in #DigitAg?

AGAP’s contributions concern methodological and software developments and their applications, especially for cereals and fruit trees.

Its research ties in with:

  •  challenges no. 1 – Agroecology (plant adaptation and contribution of plant diversity to agroecology) and no. 2 – Rapid phenotyping,
  •  axis 4 – Information systems, data storage and transfers (data flow management and reproducibility: its large quantities of data require storage servers and computing clusters, the development of databases such as DAPHNE (DAtabase PHenotype plaNt integration), and the definition of interoperability standards as well as data source management, as in the OpenAlea platform,
  •  as well as with axes 5 – Data mining, data analysis and knowledge discovery and 6 – Multiscale modelling and simulation.

Several teams from two of AGAP’s thematic fields are involved in #DigitAg:

The GE²pop team “Evolutionary Genomics and Population Management” from the thematic Diversities & Genomes: structure, domestication, environments, societies

GE²pop is an INRA-Montpellier SupAgro team that works on the evolutionary history of plant species and populations, quantifies its impact on their diversity and functioning, and develops strategies for the management and enhancement of the genetic resources of cultivated species as well as of their wild relatives. The team has strong skills in genetics and genomics, but also in bioinformatics and phenotyping. Crops concerned: durum wheat, maize and Medicago truncatula (a non-cultivated model species).

Subjects of interest for digital agriculture within #DigitAg: in phenotyping, the team calibrates spectral data (whether combined with imaging or not) collected on isolated organisms or plant cover, to quantify different leaf properties (nitrogen content, chlorophyll, specific mass). These inferences are used to specify the genetic bases of these characteristics and the effects of domestication. We also use the spectral signature to identify the contribution of different genetic components within varietal mixtures in order to specify the assembly rules and to optimise the performance of these mixtures.

With several teams involved, the thematic area Development and adaptation of plants and stands is the first concerned by digital agriculture: 

  • The AFEF team – Architecture and functioning of fruit species studies and models the genetic and ecophysiological bases of development and adaptation traits that are innovative challenges in the selection of fruit species: tree architecture, phenology and regularity of production, responses to environmental constraints (high temperatures, water stress, etc. Crops concerned: fruit trees (apple)

AGAP, AFEF Team, Proxidéeection © Cirad

Examples of subjects and tools for digital agriculture: tree development studies (foliage and fruiting) and high-throughput phenotyping of 3D architecture by LIDAR, of their summer behaviour in terms of water stress, by thermal and multispectral imaging, structure-function modelling (analysis of plant architecture that integrates plant physiological functions).

  • The PAM team – Phenotypic plasticity and adaptation of monocots analyses G (genotype) x E (environment) x C (crop management) interactions to identify traits involved in adaptation to climate and nutritional constraints, in order to steer phenotyping and integrated modelling of phenotypes and ideotypes (“ideal” varieties for a given use, product and/or environment) of plants in stands. Crops concerned: rice, sorghum, oil palm, etc.

Expertise mobilised: ecophysiology associated with histology, biochemistry, statistics and computer science.

  • The M2P2 team – Models and methods for plant phenotyping develops methods and software components to manage and analyse plant phenotyping data from automated high-throughput phenotyping platforms and from new generations of sensors (3D laser scanner, different types of cameras) that are progressively replacing manual plant phenotyping. These platforms produce spatiotemporal data and help to study development over time and to explore genetic variability and environmental effects.

Shared modelling platforms, big data processing, imaging and phenotyping

  • Coordinated by the M2P2 team, the OpenAlea software platform is a collaborative open source plant modelling platform, from the tissue level to that of the whole plant. This scientific workflow management system automatically distributes calculations, such as the analysis of phenotyping data, on distributed infrastructure (cloud, grid). The data source is retained in order to guarantee the reproducibility and traceability of results.

OpenAlea, Radiation absorption © Cirad

OpenAlea currently includes 90 models and enables their interoperability and that of modules for the acquisition, analysis and simulation of plant development.

These models concern agronomy (the impact of climate change on monospecific and mixed crops), genetics (variation in performance) and development biology (genome/environment interactions), and integrate knowledge in computer science, mathematics, physics, etc.

– A recent extension: the management and processing of very large volumes of data acquired by genotyping and phenotyping sensors and robots.
– Interfaces are being developed with other modelling platforms (CAPSIS, RECORD, GroIMP, LStudio) and the platform is part of the AMEI initiative for the interoperability of crop models.
– Examples of integrated modelling: multiscale models of fruit trees, hydraulic architecture of sorghum, modelling of mixed crops in agroecology.

  • The PHIV platform (Plant histocytology and cell imaging): develops high-throughput histology to meet the needs of anatomical phenotyping, at the tissue level. PHIV studies the role of cell and tissue components in the adaptive response of plants to their environment, their impact on agronomic traits such as root depth or the quality of biomass produced.
    • Launched in February 2019, this joint CIRAD-INRA-Montpellier SupAgro-CNRS platform proposes services and collaborations to private companies and technical institutes.
    • PHIV, associated with physicists from the L2C laboratory (University of Montpellier) in the APLIM flagship project (Advanced Plant Life Imaging and Metrology), develops magnetic resonance technologies (NMR, MRI, nanoprobes) to study plant response to abiotic and biotic constraints. An innovative portable NMR sensor has been developed to characterise water stress in plants in fields (thesis financed by LabEx Numev-CIRAD).

PHIV Platform, Imaging © Cirad

  • We can also mention the DiaPHEN platform, dedicated to phenotyping. DiaPHEN is used to conduct tests for annual crops (maize, sorghum, wheat, etc.) and fruit trees (apple) in water and heat stress conditions. It is also a place for testing and developing sensors and for collaborations with equipment suppliers. DiaPHEN is managed by the INRA DIASCOPE unit in Mauguio.

Our scientific collaborations in digital agriculture within #DigitAg

At Inra, we work with the joint research units MISTEA and LEPSE. Christophe Pradal (M2P2) is joint supervisor for the #DigitAg theses by Cyrille Midingoyi on “Semantic and modular representation of plant growth model using a declarative metalanguage” and Gaëtan Heidsieck on “Distributed management of scientific workflows for high-throughput plant phenotyping” with the University of Montpellier (Zenith team at LIRMM) and LEPSE at INRA. We also work with the EMMAH joint research unit for image analysis methodologies.

With Irstea, Jean-Luc Regnard and Magalie Delalande were involved in joint research with the TETIS research unit on image analysis of discontinuous fruit tree cover, which resulted in several co-publications with Sylvain Labbé.

Again in digital agriculture, we have a joint thesis project with LIRMM applied to fruit trees: the goal is to use machine learning methods to improve the distinction of different tree parts, in order to better estimate the number of short fruit-bearing branches.

Finally, we have implemented several projects with the company ITK: research on rice and sorghum and co-supervision of a thesis underway, also on apple trees.

What is our vision of digital agriculture?

How will a given variety behave in a given environment? This is what interests us.

For us, access to digital tools means the possibility of monitoring the behaviour of varieties in the field, whether through soil sensors, water potential sensors directly on plants, or with imaging that has a very specific role, since this methodology gives us access to the behaviour of numerous genotypes at the same time, in a particular environmental condition. Each variety reacts very specifically to its environment, and therefore spending two hours acquiring data between the first variety and the last would fail to guarantee strictly identical environmental conditions (weather fluctuations, for example). On the other hand, an aerial camera above the plant cover produces a snapshot of the field, and therefore of all genotypes at the same time, making it possible to see those under high stress and those under medium stress in the same climate conditions.

In this context, ecophysiological and statistical modelling (crop model, FSPM) also plays a key role in that it makes it possible to capture (in the form of mathematical equations) plant response processes to their environment during their life-cycle and to thereby understand and/or predict the resulting phenotypes, according to the genotype (value of model equation parameters) and the environment (temporal environmental conditions).

Thus, for data acquisition requiring rapidity, digital technologies enable instantaneous acquisition for a large number of genotypes.

Next, for statistical analysis, plant reconstruction or simulation, and image segmentation, we use software tools that need to be interconnected. The interoperability of data and analysis tools, in other words the possibility of linking them to one another, enables the automation of analysis pipelines. Further to an acquisition campaign, we work on a large number of genotypes for several days with data that is extremely heavy to manage and these pipelines are therefore essential to improve the ergonomics of analysis and to help us to automatically extract the relevant descriptors.

We also expect a great deal from deep learning to improve the quality of information extracted from data. This is an important field of research for the AGAP teams and those with which we work.

AGAP at the international level

The unit, based in Montpellier, is also present in the French overseas departments and regions and in Corsica. At the international level, we have around 15 expatriates in Africa, Latin America, Asia, Europe and Oceania, for research and training in public and private partnerships.

Two examples of international collaborations of interest to digital agriculture:

  • In the North: the goal of the CropsInsilico consortium (USA, University of Illinois), in which Christophe Pradal participates, is to establish a multiscale modelling platform, to integrate knowledge from genes to the field.
  • In the South, in Senegal, where we have been working for 20 years with CERAAS (Regional study centre for the improvement of adaptation to drought), we are providing our expertise within the Cirad incentive action dP IAVAO (Platform in partnership for research and training, Innovation and Plant Breeding in West Africa), for the implementation of a phenotyping platform based in particular on aerial imaging. The crops concerned are mainly sorghum, groundnuts, cowpeas and millet.
    • Financed by the Gates Foundation, the Sorghum Genomics Toolbox is a key project to develop tools and data (phenotyping/sequencing) with a view to accelerating sorghum breeding in the Sudan-Sahel region.

Other international partnerships are in place in South America (CIAT in Colombia), China (Shanghai University), Oceania (IAC in new Caledonia), or Mediterranean countries (Italy, Morocco, etc.). The intervention frameworks are mainly CIRAD’s platforms in partnership, but also different international centres (CGIAR) or academic establishments.

More information:

  • AGAP website
  • Contacts:
    • Direction : Patric This (Inra), Director & Emmanuel Guiderdoni  (Cirad),  Deputy Director – diragap [AT]
    • Thematic fields: Development and Adaptation of Plants and Stands: evelyne.costes [AT] / Diversities & Genomes: structure, domestication, environments, societies: xavier.perrier [AT] Integrative approaches for varietal innovation: jean-marc.lacape [AT]
    • Platforms : OpenAlea : christophe.pradal [AT] / PHIV : jean-luc.verdeil [AT]