Improving plant-based foods with AI

The global plant-based food market is currently worth around AU$80bn but is expected to grow to around AU$264bn by 2030.

Initially, companies introduced plant-based products to the market where taste and texture did not match consumer preference. Currently, the plant protein ingredients market is dominated by only a few ingredients, mostly soy and wheat-based.

Furthermore, the production processes of the existing plant protein ingredients are not optimal. The dry fractionation methods do not provide ingredients with ideal functionalities and the wet separation process has lots of potential for improvement in terms of water, energy and chemical usage. Sensory quality such as taste and texture also need to be improved.

To tackle these challenges, the Finnish RETHINK consortium, coordinated by the VTT Technical Research Centre of Finland, has kicked off a series of projects that target new ingredient technology development to boost the Finnish and global plant protein business.

Emilia Nordlund, the project leader from VTT, said, “Compared to our present highly centralised food supply chains, the project aims for improved food security by new local value chains and resilient ecosystems that can also revitalise rural areas via local farmers and industry.”

Key details of the project include:

• The project aims to create globally competitive and healthy food solutions which are based on the new scalable plant protein ingredient processes that are optimised for maximum raw material use with minimal energy and other input usage.
• The new technologies for dry fractionation and wet separation will allow the use of local plants such as oat, pea and faba bean as food ingredients, increasing security of supply.
• VTT’s approach combines machine learning with data on composition, functionality and flavour in a novel way to identify the most suitable raw materials and processes for different food applications.
• The project is expected to strengthen and increase the export by Finnish food industry by at least €500m (AU$827m).

“Intensive flavour, such as bitterness, astringency and beany or chemical-like flavour is still an issue limiting the consumption of foods manufactured with plant protein rich ingredients,” Nordlund said.

The project will look to explore processing strategies to mitigate flavour issues, such as optimising the extraction process to limit production of unfavourable compounds and reduce the content of off-flavour compounds.

One of the interesting expected outcomes of the project is creation of novel predictive models.

A methodological basis for the machine learning-based prediction of oat quality parameters was created in a previous project called OatHow, which proved the feasibility of hyperspectral imaging in identification of suitable raw materials.

“Predictive models are further developed by combining predictions on quality parameters with flavour data that will provide a competitive tool for identification of suitable raw materials for various food processes.”

Finnish industry will now work with five food companies for further exploration as part of the project.

Image credit: iStock.com/Marko Jan