Existing food processing classifications overlook the phytochemical composition of processed plant-based protein-rich foods

Abstract

According to existing food processing classification systems, plant-based protein-rich (PBPR) foods are often considered 'ultra-processed'-and therefore perceived as unhealthy-despite their ability to provide various bioactive compounds beneficial for human health. Here we used a non-targeted metabolomics approach to analyse the impact of processing on the biochemical composition of PBPR foods. Our results show that existing food classification systems may provide questionable categories for PBPR foods without considering their overall biochemical composition, including phytochemicals. An analysis focusing specifically on biochemical compounds of soy-based products manufactured using various technologies showed no clear distinctions between processing groups in the principal component analysis based on the NOVA and Poti classification. However, clear differences were found between soy-based products based on their phytochemical profile. Although food processing classification systems are welcome in their attempt to guide consumers towards healthy choices, they should be improved to more accurately reflect the biochemical composition of PBPR foods.

Fig. 1. Differences in the biochemical compositions of various food products.

a, Differences in PBPR food products and unseasoned poultry, red meat and fish products (n = 176) are shown using PCA of molecular features (n = 9,389) obtained from non-targeted LC–MS-based metabolite profiling analysis. b, Differences in PBPR food products (n = 168) are shown using PCA of molecular features (n = 9,389) obtained from non-targeted LC–MS-based metabolite profiling analysis. PC1, principal component 1; PC2, principal component 2. The variance explained by each principal component is indicated by the percentages on the x and y axes.

Fig. 2. The impact of processing on the biochemical composition of variously processed soy-based products.

a, Soy-based products classified according to product type, with PCA performed on identified compounds (n = 193). b, Soy-based products classified according to the NOVA classification system, with PCA performed on identified compounds (n = 193). c, Soy-based products classified according to the Poti et al. classification system, with PCA performed on identified compounds (n = 193). The variance explained by each principal component is indicated by the percentages on the x and y axes. d, Soy-based products classified according to product type, the NOVA classification system and the Poti et al. classification system, illustrated with a Sankey diagram. e, Identified compounds (n = 193) in soy-based products visualized using k-means clustering (z-normalized; numbers represent clusters 1–6).

Fig. 3. Relative abundances of isoflavonoids in various products processed from soybeans.

a, Relative abundances (z-normalized) of isoflavonoids in individual soy-based products expressed in a heatmap. Sample codes are shown in parentheses (Te, tempeh; E, extruded chunks; T, tofu; C, protein concentrates/isolates; B, beans). b, The differences between isoflavonoid abundances in different product types are shown. c, Total isoflavonoid abundances in three different soy-based products categorized as ultra-processed according to NOVA. Panels b and c created with BioRender.com.

Fig. 4. Diagrams illustrating soybean processing and non-targeted metabolomics analysis using LC–MS.

a, Common processing techniques used for soybeans, including the preparation of tofu, tempeh, extruded chunks, and protein concentrates or isolates. b, The non-targeted LC–MS metabolomics approach used in this study, illustrating the sample preparation, data acquisition and analysis. LC using RP and HILIC coupled with QTOF-MS with positive and negative ionizations. Figure created with BioRender.com. ESI+, electrospray ionization positive mode; ESI−, electrospray ionization negative mode.