Periodic Reporting for period 1 - E-MUSE (Complex microbial ecosystems multiscale modelling: mechanistic and data driven approaches integration.)
Reporting period: 2021-01-01 to 2022-12-31
To understand a complex microbial ecosystem and identify levers to control and/or predict its evolution, biological, mathematical, statistical and machine learning tools must be developed.
There is a bottleneck in the methodologies for extracting meaningful results from high throughput data analyses of complex biological systems, in part due to a still underdeveloped interdisciplinary interaction between biologists, mathematicians, and computer scientists. With E-MUSE, Europe will be at the forefront of these methodologies which impact many domains linked to societal and economic challenges, like novel and more controlled methods for safe and tasty food production. There can also be applications in green chemistry to predict the evolution of microbial ecosystems for bioconversion processes or in sustainable development for bio-compounds production.
The E-MUSE training programme aims at developing young researchers’ skills at the interface between artificial intelligence and life sciences. The research programme of the E-MUSE network is to develop innovative methodologies that will be applied to model structural and dynamic features of microbial communities, to identify key processes and biomarkers for specific applications.
The macroscopic properties of a fermented product emerge from the action of microbial communities which in turn depends on the environmental conditions. To be able to predict the emerging properties, innovative modelling techniques, combining mechanistic and data-driven approaches, must be developed to connect the scales and include the dynamics of the microbial communities. The early-stage researchers (ESRs) with biology, mathematics or modelling projects are interacting to understand the structure and function of microbial communities and predict their evolution.
There is a bottleneck in the methodologies for extracting meaningful results from high throughput data analyses of complex biological systems, in part due to a still underdeveloped interdisciplinary interaction between biologists, mathematicians, and computer scientists. With E-MUSE, Europe will be at the forefront of these methodologies which impact many domains linked to societal and economic challenges, like novel and more controlled methods for safe and tasty food production. There can also be applications in green chemistry to predict the evolution of microbial ecosystems for bioconversion processes or in sustainable development for bio-compounds production.
The E-MUSE training programme aims at developing young researchers’ skills at the interface between artificial intelligence and life sciences. The research programme of the E-MUSE network is to develop innovative methodologies that will be applied to model structural and dynamic features of microbial communities, to identify key processes and biomarkers for specific applications.
The macroscopic properties of a fermented product emerge from the action of microbial communities which in turn depends on the environmental conditions. To be able to predict the emerging properties, innovative modelling techniques, combining mechanistic and data-driven approaches, must be developed to connect the scales and include the dynamics of the microbial communities. The early-stage researchers (ESRs) with biology, mathematics or modelling projects are interacting to understand the structure and function of microbial communities and predict their evolution.
ESR1 is exploring and implementing new approaches to predict the optimal metabolic behaviour for mutualistic microbial consortia that experience periodic changes in their environments. The first simulations carried with this approach have been presented in a poster at the conference “The future of physics of life”, organized by AMOLF in Amsterdam in June 2022.
ESR2 started to work on the modelling at two different scales: the ecological scale and the genome scale. ESR2 undertook a first approach consisting of the use of an ecological model to identify the extra-cellular interactions occurring within the community. The modelling at the genome scale focuses on the reconstruction of the metabolic network of a yeast involved in the cheese ripening process.
ESR3 has implemented the Partial Integral Differential Equation model that captures the stochastic dynamics of the gene regulatory networks. This implementation has been published in IFAC-PapersOnLine in 2022. ESR3 is developing optimal control and model predictive algorithms to reach a desired state. This work will be presented at "Gordon Research Conference" for "Synthetic biology" in 2023.
ESR4 is collecting data on one of the lactic bacteria used during the ripening of semi-hard cheese such as Gouda and Cheddar. The reconstruction and curation of the genome-scale metabolic model undertaken will contribute to the characterization of the species.
ESR5 is studying the role of iron in the growth, interaction and production of volatile flavour compounds of cheese-ripening microorganisms, namely bacteria from Livarot-type smear-ripened cheeses. The predictive growth model was developed for 5 ripening bacteria. To investigate their interaction, genome scale metabolic models were reconstructed for two selected bacteria.
ESR6 is exploring kernel versions of dimensionality reduction methods. ESR6 has proposed a new feature selection method to improve the interpretability of the kernel version of principal component analysis. This method was presented at the Statistical Methods for Post Genomic Data Conference in February 2023.
ESR7 is studying inter and intra-species interactions using network theory approaches. The protein network related to a reduced cheese microbial community has been reconstructed. In parallel to this reconstruction, ESR7 has obtained promising results for the prediction of metabolites amount in cheese from microbial expression profiles.
ESR8 is exploring classification methods, namely SuperTML, a method embedding tabular data into a 2d representation. ESR8 has created a representation similar to the SuperTML one and tested these representations on different benchmarks. The first results have shown the good performance of the method.
ESR9 is developing deep learning methods for unveiling the relationships between the omics data coming from the cheese production process and the resulting product. ESR9 has tested deep learning methods and has implemented a workflow improving their reproducibility.
ESR10 aims to understand the effect of iron on a model cheese ecosystem using a reduced microbial community of 9 species. First experiments have been implemented to set up the conditions of iron sources and concentrations. Milk-based and plant-based fermented products have been tested. ESR10 presented the first results in a poster at FoodMicro 2022 Conference in Athens.
ESR11 is working on the predictive modelling methodologies that examine the biological, physical, and chemical mechanisms underlying the cheese ecosystems, and aims to interpret the complexities of microbial communities with two computational approaches: Individual-based Models and Partial Differential Equations. The first implementation of an individual-based model aimed to describe the spatial dynamics of microorganisms.
ESR12 is implementing a comprehensive study of Maltese Gbejna cheeses focusing on microbiota identification by combining both molecular and morphological methods and assessing their impact on cheese’s properties.
ESR13 has set up a high-throughput screening system and fermentations with 33 lactic acid bacteria strains were conducted in 6 different plant-based ingredients. Acidifying capacities were measured and aromatic volatile compounds produced were analysed by a semi-quantitative method. ESR13 presented her results at FoodMicro 2022 Conference in Athens.
ESR14 has analysed a dataset on the ripening of Gouda-type cheese in a micro cheese model system. ESR14 investigated the effects of temperature modifications at different cheesemaking stages and ripening time on cheese flavour development.
ESR15 has developed a mechanistic model to predict the dynamics of phage attack during milk acidification. The model is able to predict satisfactorily most of the cases. These results were presented at EFFoST conference in 2022 in Dublin.
ESR2 started to work on the modelling at two different scales: the ecological scale and the genome scale. ESR2 undertook a first approach consisting of the use of an ecological model to identify the extra-cellular interactions occurring within the community. The modelling at the genome scale focuses on the reconstruction of the metabolic network of a yeast involved in the cheese ripening process.
ESR3 has implemented the Partial Integral Differential Equation model that captures the stochastic dynamics of the gene regulatory networks. This implementation has been published in IFAC-PapersOnLine in 2022. ESR3 is developing optimal control and model predictive algorithms to reach a desired state. This work will be presented at "Gordon Research Conference" for "Synthetic biology" in 2023.
ESR4 is collecting data on one of the lactic bacteria used during the ripening of semi-hard cheese such as Gouda and Cheddar. The reconstruction and curation of the genome-scale metabolic model undertaken will contribute to the characterization of the species.
ESR5 is studying the role of iron in the growth, interaction and production of volatile flavour compounds of cheese-ripening microorganisms, namely bacteria from Livarot-type smear-ripened cheeses. The predictive growth model was developed for 5 ripening bacteria. To investigate their interaction, genome scale metabolic models were reconstructed for two selected bacteria.
ESR6 is exploring kernel versions of dimensionality reduction methods. ESR6 has proposed a new feature selection method to improve the interpretability of the kernel version of principal component analysis. This method was presented at the Statistical Methods for Post Genomic Data Conference in February 2023.
ESR7 is studying inter and intra-species interactions using network theory approaches. The protein network related to a reduced cheese microbial community has been reconstructed. In parallel to this reconstruction, ESR7 has obtained promising results for the prediction of metabolites amount in cheese from microbial expression profiles.
ESR8 is exploring classification methods, namely SuperTML, a method embedding tabular data into a 2d representation. ESR8 has created a representation similar to the SuperTML one and tested these representations on different benchmarks. The first results have shown the good performance of the method.
ESR9 is developing deep learning methods for unveiling the relationships between the omics data coming from the cheese production process and the resulting product. ESR9 has tested deep learning methods and has implemented a workflow improving their reproducibility.
ESR10 aims to understand the effect of iron on a model cheese ecosystem using a reduced microbial community of 9 species. First experiments have been implemented to set up the conditions of iron sources and concentrations. Milk-based and plant-based fermented products have been tested. ESR10 presented the first results in a poster at FoodMicro 2022 Conference in Athens.
ESR11 is working on the predictive modelling methodologies that examine the biological, physical, and chemical mechanisms underlying the cheese ecosystems, and aims to interpret the complexities of microbial communities with two computational approaches: Individual-based Models and Partial Differential Equations. The first implementation of an individual-based model aimed to describe the spatial dynamics of microorganisms.
ESR12 is implementing a comprehensive study of Maltese Gbejna cheeses focusing on microbiota identification by combining both molecular and morphological methods and assessing their impact on cheese’s properties.
ESR13 has set up a high-throughput screening system and fermentations with 33 lactic acid bacteria strains were conducted in 6 different plant-based ingredients. Acidifying capacities were measured and aromatic volatile compounds produced were analysed by a semi-quantitative method. ESR13 presented her results at FoodMicro 2022 Conference in Athens.
ESR14 has analysed a dataset on the ripening of Gouda-type cheese in a micro cheese model system. ESR14 investigated the effects of temperature modifications at different cheesemaking stages and ripening time on cheese flavour development.
ESR15 has developed a mechanistic model to predict the dynamics of phage attack during milk acidification. The model is able to predict satisfactorily most of the cases. These results were presented at EFFoST conference in 2022 in Dublin.
Interactions between ESRs from different disciplines have already shown the added value of the interdisciplinary E-MUSE project to develop new methodologies. Knowledge on several microorganisms and cheese types have already been collected, new methodologies to extract features from large datasets and new modelling approaches have been implemented and tested. The collaboration between ESRs to develop and apply these innovative methodologies will give a comprehensive view of cheese microbial ecosystem. Modelling methods that will be developed will allow the prediction of cheese properties from micro-scale information. The implementation of these predictive models could guide the cheese makers to avoid spoilage of production and to produce cheese with qualities favoured by the consumers.