Extended Data Fig.9: CAZyme-associated plasma proteome responses to consumption of the 4-fibre snack prototype.

a-c, Contextual language processing literature analysis (CompBio) of proteins whose abundances were significantly correlated with treatment-discriminatory CAZymes in participants consuming the 4-fibre snack. The analysis procedure is summarized in Extended Data Fig.5. Treatment-responsive proteins, identified by CC-SVD as having projections at the extremes of singular vector 1 (10^th and 90^th percentiles of the distribution), are grouped into biological themes identified by CompBio, based on a conditional probability analysis, as being significantly enriched for contextually-associated biological ‘concepts’ (processes/pathways) over those that occur by random sampling of the literature. Themes with enrichment scores (log₂) > 14.8 in the plasma proteomes of participants who consumed the 4-fibre snack are shown in panels b and c (see Supplementary Table10d-f for a comprehensive list of themes associated with this and the other fibre snacks). d,e, Biological themes based on proteins positioned in 10^th and 90^th percentiles (panels d and e, respectively) are portrayed as spheres. The size of a sphere is related to its enrichment score in the plasma proteome after consumption of the 4-fibre snack. The thickness of the blue lines connecting themes signifies the number of proteins shared between them. Component proteins of exemplary themes (orange spheres) are listed in boxes and coloured by their median log₂ fold-change in response to consumption of the snack (blue, decrease; red, increase).