Simulating interactions in microbial communities through Physics Informed Neural Networks: towards interaction estimation

Paguiel Javan Hossie (1) , Béatrice Laroche (2) , Thibault Malou (2) , Lucas Perrin (3) , Thomas Saigre (4, 5) , Lorenzo Sala (2)

1 IDP - Institut Denis Poisson
2 MaIAGE - Mathématiques et Informatique Appliquées du Génome à l'Environnement [Jouy-En-Josas]
3 ANGE - Numerical Analysis, Geophysics and Ecology
4 IRMA - Institut de Recherche Mathématique Avancée
5 Cemosis

Microorganisms form complex communities known as microbiota, influencing various aspects of host well-being. The Generalized Lotka-Volterra (GLV) model is commonly used to understand microorganism population dynamics, but its application to the microbiota faces challenges due to limited bacterial data and complex interactions. This preliminary work focuses on using a Physics Informed Neural Network (PINN) and synthetic data to simulate bacterial species evolution driven by a GLV model. The approach is calibrated and tested on several models differing in size and dynamic behavior.

Mots clés: Physics Informed Neural Networks; System of ODEs; Generalised Lotka Volterra model; Population dynamics; Microbiota

Domaines: Réseau de neurones [cs.NE]; Systèmes dynamiques [math.DS]; Machine Learning [stat.ML]

Paguiel Javan Hossie, Béatrice Laroche, Thibault Malou, Lucas Perrin, Thomas Saigre, et al.. Simulating interactions in microbial communities through Physics Informed Neural Networks: towards interaction estimation. 2024. ⟨hal-04440736⟩