Periodic Reporting for period 2 - OptiSignFood (Data Science and AI assisted holistic software to digitally design optimised high quality and safe food products with minor environmental impact)
Reporting period: 2023-01-01 to 2024-06-30
New food product development is still an inefficient, time-intensive and manual process based on an iterative laboratory approach. In the food design process, a multitude of parameter, e.g. consumer’s wishes, food safety, health requirements and environmental related needs, must be considered. It takes up to 24 months and 150 tests to create a new food product. We will overcome this challenge with a disruptive digital food design solution based on data science and AI. The OptiSignFood software will provide a holistic and integrated approach to predict and optimise food’s quality, appearance and safety based on reliable scientific data. OptiSignFood will be the first solution on the market able taking in consideration all aspects of food design. The AI based approach allows 85% more accurate predictions than competitor's simple calculations. It saves manufacturers 90% of time in the food product design and helps achieving Green Deal and UN sustainable development goals by creating more sustainable and healthy food products selecting ingredients with same nutritional properties, but lower environmental impact, saving up to 76% CO2eq.
The translational consortium combines nutrition, food technology and environmental expertise from the research and industry to foster the transfer of the innovative solution from the lab into a product that meets the customer needs:
The Makers Food (Berlin), a modern FMCG company using data-science and AI to produce better food with lowest environmental footprint and optimised nutritional values. The Makers Food team laid the foundation for OptiSignFood and developed the AI prediction models and the platform.
Agroscope (Zurich) is one of the leading institutions for sustainability in the food chain and focuses on the concept to connect environmental and nutrition/health impacts.
Metacognis (London) is specialised on life science and AI data mining.
Pascal Processing (Helmond, NL) is a contract manufacturer and will contribute real data to the project and become the first industry user.
The translational consortium combines nutrition, food technology and environmental expertise from the research and industry to foster the transfer of the innovative solution from the lab into a product that meets the customer needs:
The Makers Food (Berlin), a modern FMCG company using data-science and AI to produce better food with lowest environmental footprint and optimised nutritional values. The Makers Food team laid the foundation for OptiSignFood and developed the AI prediction models and the platform.
Agroscope (Zurich) is one of the leading institutions for sustainability in the food chain and focuses on the concept to connect environmental and nutrition/health impacts.
Metacognis (London) is specialised on life science and AI data mining.
Pascal Processing (Helmond, NL) is a contract manufacturer and will contribute real data to the project and become the first industry user.
A major part of the work carried out during the project was to acquire the missing data, reformatting, harmonising and fixing them where necessary and developing a meta-database for the OptiSignFood platform, but also to improve the prediction performance.
Data were used from own measurements, but also from existing databases such as EuroFIR (The European Food Information Resource Network project), which includes more than 20 national and specialised food sub-databases. Within the project, six country-specific food composition databases were licensed from EuroFIR to increase the relevance for the European market and the number of foods covered. Pre-processing was necessary to harmonise and standardise the data. The sub-databases were then merged into an aggregated nutrition database. For the environment database, the Life Cycle Inventory (LCI) data of food ingredients from five different databases were prepared, harmonised and standardised for integration into the meta-database.
Separate datasets on pH, colour and texture parameters such as Bostwick Consistency, collected by The Makers Food and Pascal Processing, were combined into an aggregated food technology database containing information on more than 300 ingredients.
In addition, Metacognis used its Heron data mining application to increase the amount of data available for OptiSignFood, thereby improving its accuracy.
The aggregated databases, the nutrition database, the environment database and the food technology database were used to create the meta-database platform.
By increasing the size and number of databases, algorithms were developed to calculate and predict relevant food parameters. The models for nutrition, pH, colour and texture were improved compared to the prototype or preliminary state of 2020. The implementation of the Life Cycle Inventory databases enabled the calculation of specific environmental indicators. The first step was to select indicators that are widely used. Once the framework is established, additional life cycle indicators can be easily integrated.
In parallel, a secure development lifecycle design and user research were carried out to identify the current pain points of the users, but also to get their feedback on the initial UX/UI design.
In the second reporting period, the focus was on creating an OptiSignFood software application for demonstration and validation by pilot customers. An initial user manual was developed, the UX/UI was further improved, and the frontend and backend software was implemented. User needs were identified and prioritised to create an MVP. The models' performance was evaluated with test and industry data sets. 21 pilot customers tested the OptiSignFood platform with overall positive feedback. This feedback was used to identify opportunities to improve usability, interface design and functionality.
Data were used from own measurements, but also from existing databases such as EuroFIR (The European Food Information Resource Network project), which includes more than 20 national and specialised food sub-databases. Within the project, six country-specific food composition databases were licensed from EuroFIR to increase the relevance for the European market and the number of foods covered. Pre-processing was necessary to harmonise and standardise the data. The sub-databases were then merged into an aggregated nutrition database. For the environment database, the Life Cycle Inventory (LCI) data of food ingredients from five different databases were prepared, harmonised and standardised for integration into the meta-database.
Separate datasets on pH, colour and texture parameters such as Bostwick Consistency, collected by The Makers Food and Pascal Processing, were combined into an aggregated food technology database containing information on more than 300 ingredients.
In addition, Metacognis used its Heron data mining application to increase the amount of data available for OptiSignFood, thereby improving its accuracy.
The aggregated databases, the nutrition database, the environment database and the food technology database were used to create the meta-database platform.
By increasing the size and number of databases, algorithms were developed to calculate and predict relevant food parameters. The models for nutrition, pH, colour and texture were improved compared to the prototype or preliminary state of 2020. The implementation of the Life Cycle Inventory databases enabled the calculation of specific environmental indicators. The first step was to select indicators that are widely used. Once the framework is established, additional life cycle indicators can be easily integrated.
In parallel, a secure development lifecycle design and user research were carried out to identify the current pain points of the users, but also to get their feedback on the initial UX/UI design.
In the second reporting period, the focus was on creating an OptiSignFood software application for demonstration and validation by pilot customers. An initial user manual was developed, the UX/UI was further improved, and the frontend and backend software was implemented. User needs were identified and prioritised to create an MVP. The models' performance was evaluated with test and industry data sets. 21 pilot customers tested the OptiSignFood platform with overall positive feedback. This feedback was used to identify opportunities to improve usability, interface design and functionality.
OptiSignFood is a data science and AI assisted holistic software to digitally design optimised high quality and safe food products with minor environmental impact. Our AI digital food design process can safe valuable time in the recipe formulation and validation (simple flavour pairing solution already proved 70% time savings). OptiSignFood will be the first integrated solution available on the market that will couple Health, Safety and Environmental features to design new Sustainable Food Products.
Starting from a selection of ingredients OptiSignFood is expected to:
+ perform a multivariate optimisation of food quality parameters, i.e. nutrients, colour and texture
+ predict the food safety characteristic (i.e. pH value and microbiological inactivation) according to composition and diverse food processing approaches (e.g. high-pressure processing, thermal preservation, fermentation),
+ predict the cradle-to-gate footprint of the final product.
On top, as a result of the data modelling algorithms, OptiSignFood will perform a reverse optimisation of the recipe from a set of final characteristics and properties. This means that our clients save up to 90% of time in the new food product development process. Moreover, the cradle-to-gate environmental impact of a food product can be optimised substituting an ingredient with a high environmental footprint by one with a lower one, maintaining the same quality and safety characteristics of the final product. This can lead to up to 98% CO2eq savings in food products with high protein content (considering daily protein intake). Because OptiSignFood considers both nutritional aspects and environmental impacts, we are able to develop a methodology for a score that informs brands, producers and consequently consumers about food that is on a good middle ground, not sacrificing health for sustainability or the other way around.
Starting from a selection of ingredients OptiSignFood is expected to:
+ perform a multivariate optimisation of food quality parameters, i.e. nutrients, colour and texture
+ predict the food safety characteristic (i.e. pH value and microbiological inactivation) according to composition and diverse food processing approaches (e.g. high-pressure processing, thermal preservation, fermentation),
+ predict the cradle-to-gate footprint of the final product.
On top, as a result of the data modelling algorithms, OptiSignFood will perform a reverse optimisation of the recipe from a set of final characteristics and properties. This means that our clients save up to 90% of time in the new food product development process. Moreover, the cradle-to-gate environmental impact of a food product can be optimised substituting an ingredient with a high environmental footprint by one with a lower one, maintaining the same quality and safety characteristics of the final product. This can lead to up to 98% CO2eq savings in food products with high protein content (considering daily protein intake). Because OptiSignFood considers both nutritional aspects and environmental impacts, we are able to develop a methodology for a score that informs brands, producers and consequently consumers about food that is on a good middle ground, not sacrificing health for sustainability or the other way around.