There is a strong need for better sensors to guide the food industry transition to plant-based products. These will allow the discovery, monitoring and prediction of both positive and negative sensory qualities of novel food products and ingredients. The industry struggles with the ethical and financial burden of using human taste panels to identify novel ingredients from novel sources. Also, they seek more efficient ways to understand and eliminate off-flavors and to identify and formulate healthy and flavorful novel ingredients.
In that light the recently published microfluidic Tongue-on-a-Chip platform developed by Wageningen Plant Research (Roelse 2024) could have a large impact. The paper demonstrates that we can measure a variety of taste attributes based on a human taste receptor array, blocking or enhancing of taste responses by matrix components (bitter off-tastes), and even sensory kinetic aspects such as onset and lingering of sweeteners. We, thus, believe the Tongue-on-a-Chip technology combines the best of two worlds: quantitative high content measurements using multiple human taste receptor assays, which also produce kinetic sensory parameters that correlate to human panel ratings. Now that the feasibility has been demonstrated, this project aims to bring the technology further towards application (TRL level 5-6)
The current platform is bioluminescence-based, and allows the measurement of complex colored, fluorescent or non-transparent samples. However, there are some factors in foods that induce a generic host cell response (HCR) via endogenously expressed receptors. These background signals are complex to separate from the receptor-specific ones and require the endogenous receptors to be inactivated. Secondly, for flavor it is desirable to also assess aroma (nose), and, thirdly, there is a need for more accessible tools (algorithms) to interpret the kinetics of taste and aroma receptor response profiles (fingerprints) and correlate them to human sensory panels to provide valid flavor/health predictions.
We identified the main endogenous HCR receptors and already successfully knocked out by gene editing a first set of five receptors in a single cell line for immediate evaluation in this project. For any remaining HCR issues an additional set of endogenous receptor targets has already been identified. Next, the plan is to add an aroma analysis hardware component to the platform to enable aroma profiling and intensity analysis with olfactory receptors. With this improved bio- and hardware we will subsequently implement machine learning algorithms by correlating the kinetic taste and aroma receptor response profiles to human panel ratings, offered by the companies, for accurate flavor predictions by the SENSORY-ON-CHIP 2.0. This validation will ultimately trigger a wide application perspective for the food industry and beyond.

The technical objectives of this Innovation Program ST1 project are therefore threefold: Firstly, sensor bioware for clean signals: implement editing knock outs of multiple genes to eliminate host cell response (HCR) interference when using our cell lines. Secondly, sensor hardware to implement aroma analysis: an aroma-exchanger based on microbubbles to test aromatic headspace of food samples against olfactory receptor cell arrays. Thirdly, trained machine learning ML
algortihms which take features of sample series from multiple experiments and correlates them to human panel flavor ratings in a first step towards a human sensory digital twin. This 3-pronged approach will overcome the current limitations of the technique allowing analyses and predictions at end-product concentrations and unlock the full potential for food and health applications of both receptomics and conventional titerplate based receptor assays.

The project develops prototypes and demonstrators with novel bioware (sensor cell lines) and hardware (novel component for inline aroma detection) jointly generating big data to be integrated into trainable software (ML) and a digital twin of flavor for quality predictions as part of Innovation Program ST1 Enabling Smart Technology and serving Mission 4 on Sustainable and valued food that is healthy, accessible and safe, especially sections 4C Alternative proteins and 4D Sustainable and healthy food products that are appreciated by the consumers because they simply taste better with off tastes removed or blocked and desirable flavors and healthy ingredients added based on insights delivered by this technology.

Concrete results of the project for which partners/users?
• A tongue-on-chip 2.0: human host cell lines suitable for testing complex undiluted (= full flavor)
extracts against any choice array of receptors for flavor applications. Very broad use in the human
(plant) food flavor and animal feed field but also beyond for environmental (pollution detection) and
pharmaceutical (biomedicines) applications
• A nose-on-chip 2.0: a novel component for inline aroma-to-medium exchange to allow direct seamless
analysis of headspace samples vs arrays of olfactory receptors in the current platform. This will
provide fingerprint information on aroma differences and intensities.
• SENSORY-ON-CHIP 2.0: machine learning algorithms to correlate the variation in receptor responses
with variation in human panel ratings of flavor sample series aspects and perform accurate predictions
with novel sample sets. Can be used to predict flavor ratings during product formulation and as such
can be seen as the beginning of a digital twin of human sensory experience.
• An instrument & software prototype which in principle can be introduced and implemented in the
partner labs and introduced elsewhere via our ReceptomiX partner.