Fig. 4: Abiraterone acetate (AA) exerts a reproducible effect on the human gut microbiota in vivo and in vitro.

a Differentially abundant pathways in patients receiving AA compared to those who were not. Multiclass analysis with the LEfSe algorithm was used to discriminate between AA effects overlapping systemic ADT exposure (LDA score >2 and p < 0.05 for all pathways shown). b–e Relative abundance of predicted pathways that were statistically increased in AA-treated patients. Data represent the median (line in box), IQR (box), and minimum/maximum (whiskers) of CTRL (n = 33), ADT (n = 21), and ADT + AA (n = 14) patient samples. Statistics shown for Kruskal–Wallis tests with multiple comparisons corrected using the Benjamini–Hochberg FDR method. f–i Relative abundance of predicted pathways in AA-exposed gut model samples. j Relevant bacterial contribution to each menaquinone biosynthesis pathway. Predicted pathways were inferenced using an exact amplicon sequence variant approach with the PICRUSt2 software and annotated using the MetaCyc metabolic pathway database. n.s. = not significant.