Extended Data Fig. 1: Design of the defined bacterial consortium RCom.
a, Schematic of the design process. The key criteria for selecting the 15 bacterial species for the RCom are shown in colored boxes. Based on metagenomic shotgun sequencing of faecal samples from anti-PD-1-treated NSCLC patients, the bacterial species enriched (LDA score ≥ 3.0) and prevalent (prevalence = 1.0) in the responder group were identified. For safety, the species associated with ≥grade 3 irAEs or harboring putative toxins or multiple drug-resistant genes were excluded. Lastly, a metabolically cooperative community was predicted based on metabolic interaction simulations. b, Selection of bacterial species based on enrichment and prevalence in responders and safety rule. The bacterial species that meet the criteria are shown in green boxes, whereas those excluded are denoted by bold X. c, Metabolic model-based prediction of interspecies interactions in bacterial communities. Calculated competition (MRO) and cooperation (SMETANA) scores for different sizes of bacterial communities are shown on the right. The SMETANA score was normalized by community size. Arrow indicates the RCom. d, Number of predicted intra- and inter-phylum cross-feeding interactions within the 11 core members and the RCom. e, Metabolite classes predicted to be exchanged between the RCom members. Compound classification is based on the Human Metabolome Database (HMDB). f, Top 20 potential cross-fed compounds in the RCom. Amino acids are indicated in brown.
