Mots-Clés
bioinformatics
PINNs
GRN
gene regulatory networks
plant-pathogens interaction
hybrid models
tomato
Description
Leveraging scientific machine learning to decode the molecular response of plants exposed to multiple stresses.
Topic:
A major question in plant biology is to understand how plant growth, development, and environmental responses are controlled and coordinated by the activities of regulatory factors. In particular, infection triggers a dynamic cascade of reciprocal events between host and pathogen wherein the host activates complex mechanisms to recognise and kill pathogens while the pathogen often adjusts its virulence and fitness to avoid eradication by the host. The interaction between the pathogen and the host results in large-scale changes in gene expression. Timely and rapid plant response to these attacks is essential and can dramatically affects plants fate. Uncovering how transcription factors (TFs) regulate their targets at different molecular levels over time is critical to define gene regulatory networks (GRNs) in normal and diseased states. While several different types of high-throughput experimental procedures are available to study systems in the cell, most only measure static properties of such networks. Despite decades of advancement, challenges remain in GRN inference, including dynamic rewiring.
With this project we propose to develop a hybrid model at the crossroad between mathematics and artificial intelligence to study how the complex GRN in plants is dynamically modulated to during pathogens’ attack.
Because of the large diversity of hazards and its important agroeconomic interest, we will focus on tomato (Solanum lycopersicum) as biological model. In Dr. Bottini’s team, we have selected and collected high-quality omics data publicly available, which are organized in an internal FAIR database, POMOdOROO (Pan OMics cOllection of tOmato undeR biOtic and abiOtic stress). Another important resource developed in our team is TomTom, a knowledge graph for tomato gathering 11 databases which represents a fingerprint of a wide type of molecule interactions comprising 113 415 entities and 2 864 036 relationships. Leveraging those two databases, the needed omics data and the biological knowledge to include in the model are already available and ready to be used for the model development. Lately, we have been working to set up the basis of a hybrid model to model the dynamical changes of gene expression in tomato upon pathogen attack without considering the interactions among different genes. Here, we wish to upgrade this initial model to study how the complex GRN is rewired in diseases versus non-diseased plants and in response to different stresses.
Here’s the detailed tasks:
Task 1.1 Define the mathematical equations governing the dynamics of gene regulatory networks based on the Hill function which has been used in the literature to model the regulation of gene expression and our preliminary results.
Task 1.2 Building the GRN skeleton based on curated prior knowledge from TomTom and omics time-series data from POMOdOROO.
Task 1.3 Implementation of the BINN algorithm using the biological knowledge from task 1.2 and the physics from task 1.1 and application on data from POMOdOROO.
We wish to obtain at the end of the internship a novel model which combines the power of AI for modeling complex systems, of biological knowledge for the mechanistic explication and of the mathematical model for the predictive ability, to surpass the scale limitations of traditional modeling methods while being fully interpretable.
Tutor:
Silvia Bottini, Junior Professor Chair INRAe/UniCA – Team leader; team SMILE at the Institut Sophia Agrobiotech, silvia.bottini@inrae.fr
Régis Duvigneau, DR Inria; team ACUMES, Inria Sophia-Antipolis, regis.duvigneau@inria.fr
Where:
Institut Sophia Agrobiotech, Sophia-Antipolis: https://institut-sophia-agrobiotech.paca.hub.inrae.fr/equipes-isa/smile
References
Multari M, Carriere M, Amoros-Gabarron X, Damy A, Lobentanzer S, Saez-Rodriguez J, Jaubert S, Dugourd A, Bottini S. A knowledge graph and topological data analysis framework to disentangle the tomato-multi pathogens complex gene regulatory network. Available from: https://www.biorxiv.org/content/10.1101/2025.04.09.647963v1
Calia G, Marguerit S, Mota APZ, Vidal M, Schuler H, Brasileiro ACM, Guimares P, Bottini S. Modelling single-stress omics integration with HIVE enables the identification of responding signatures to multifactorial stress combinations in plants. Available from: https://www.biorxiv.org/content/10.1101/2024.03.04.583290v3
Vidal M, Duvigneau R, Bottini S Modélisation de la dynamique de l’expression des gènes des plantes lors d’une attaque biotique par intelligence artificielle informée par la physique. Available from : https://inria.hal.science/INRIA-RRRT/hal-05210203v1