By combining 3D imaging and artificial intelligence, two recent scientific publications have opened a new path to design protective biofilms.
These microbial communities, when properly assembled, can block pathogens such as Salmonella. This innovative experimental and predictive approach reduces reliance on antibiotics and biocides, with promising applications for animal health, food safety, and environmental protection.
Bacteria rarely live in isolation. Instead, they organise into complex 3D communities attached to surfaces, called biofilms. Ubiquitous in natural and agricultural environments, these structures can either promote pathogen persistence or act as natural protective barriers. However, identifying and assembling biofilms that actively protect against harmful microorganisms remains a major scientific and technological challenge.
Within the framework of the ANR LabCom Biofilm1Health, researchers from INRAE, the company Lallemand, AgroParisTech, and ANSES have joined forces to explore innovative strategies to control microbial interactions. Their approach is consistent with the One Health concept, which aims to simultaneously protect human, animal, and environmental health while limiting the use of antibiotics and biocides. This work paves the way for natural and sustainable biosecurity solutions for livestock farming and the food industry.
Main findings
The first study, published in ISME Communications, demonstrates through 3D and 4D confocal imaging that certain bacterial combinations,particularly Bacillus velezensis and Pediococcus, form protective biofilms capable of efficiently excluding pathogens such as Escherichia coli, Staphylococcus aureus, Enterococcus cecorum, and Salmonella enterica. These biofilms act through competition for nutrients and spatial niches, as well as more complex ecological interactions modelled mathematically using Jameson and Lotka–Volterra equations.
The second study, published in Artificial Intelligence in the Life Sciences, builds on these datasets by applying machine learning. The researchers developed a predictive model capable of estimating whether a beneficial strain can exclude a pathogen based solely on morphological descriptors of single-species biofilms (volume, thickness, roughness, etc.). This approach drastically reduces the need for exhaustive experimental testing and highlights key parameters governing microbial interactions.
By bridging advanced imaging and AI-based modelling, this research introduces a novel predictive strategy to design protective biofilms, offering direct applications in animal health, food safety, and environmental sustainability.
Contact:
Links:
Guéneau V, Guillier L, Berdous C, Noirot-Gros MF, Jimenez G, Plateau-Gonthier J, Serror P, Castex M, Briandet R. 3D imaging-driven assembly of multispecies biofilms with antagonistic activity against undesirable bacteria. ISME Commun. 2025 Sept5:5(1).
Rubrice R, Gueneau V, Briandet R, Curnuejols, Guigue V. A machine learning framework for the prediction and analysis of bacterial antagonism in biofilms using morphological descriptors. Artificial Intelligence in the Life Sciences. Vol. 8, December 2025, 100137.
