An Artificial Intelligence approach towards screening of the food matrix and digestion effects in predicting phytochemical bioaccessibility

Abstract

Functional foods are recognised to confer additional health benefits other than basic nutrition. These desirable characteristics can be attributed to specific bioactive compounds, such as vitamins or polyphenols. While the topic of health claims motivates extensive research, there is a notable gap in addressing how the carrier functional food will impact the gastrointestinal journey of a substance of interest. This consideration is crucial, as suboptimal food designs can lead to a compound being degraded or excreted instead of absorbed in a usable form, in which case the post-consumption effects are negligible. The underlying physicochemical mechanisms occurring during digestion are complex. Previous studies provide conceptual knowledge regarding what factors impact a compound’s absorption, but there is a need for methodologies that translate this understanding into practical decision-support systems. In this thesis, a systematic approach for the study of a compound’s absorption as a function of its carrier food’s properties is presented, based on systematic and empirical screening experiments. The subjects of study were fat-soluble compounds, for which it is established that the main limiting step in absorption is the release from the food and solubilisation in the digesta (bioaccessibility). In the case of vitamin E, high and invariable bioaccessibilities were observed across all formulations (~85%), but curcuminoids presented a wide variability of low and moderate bioaccessibilities (~12%-~55%) which were investigated further. The results indicated that variations in macronutrient concentration, as well as other factors, presented strong associations to this variability (p-value < 0.01). These insights were used to develop a proof-of-concept mathematical model capable of generalization. Its preliminary estimations closely resembled the observations (<10% relative error), setting high expectations for future studies.