Fig. 5: Connecting taxonomic composition to CAZyme repertoires in HIS and LMIS microbiomes.
From: Cayman enables large-scale analysis of gut microbiome carbohydrate-active enzyme repertoires
a, Heat map showing predicted taxonomic contributions to CAZyme abundances in HIS (n = 3,166) and LMIS (n = 794) gut microbial communities. Coloured cells depict the R2 value of multivariable linear models in which log10-scaled CAZyme abundance values were predicted from all CLR-transformed species (mOTU) abundances within a given genus. Blue and red cells indicate better fits in HIS or LMIS models, respectively, while purple cells indicate similar fits for HIS and LMIS models (R2 values within twofold difference). Grey cells indicate models with an insignificant fit after multiple testing correction at 1% FDR. HIS versus LMIS enrichments for taxa and CAZy families were defined as previously (Fig. 4) and substrate annotations for CAZymes are included as the left-most columns. b–g, Scatter plots between taxonomic abundances of different genera and abundances of different CAZymes: between Bifidobacterium abundance and GH13_30 family abundance (b), between Bifidobacterium abundance and GH13_3 family abundance (c), between Collinsella abundance and GH13_30 family abundance (d), between Bacteroides abundance and GH95 family abundance (e), between Prevotella abundance and GH95 family abundance (f), and between Akkermansia abundance and GT31 family abundance (g). In all of these cases, genus- and mOTU-level associations with CAZyme abundance were similar (Extended Data Fig. 9). Linear fits are indicated by R2 values (Methods).
