Fig. 1

Construction and assessment of the pandaGUT database quality and representation. A Geographic distribution of the 439 giant panda stool samples collected from across all three wild genetic populations and most captive populations, with individuals spanning different dietary regimes, ages, sexes, and geographic distributions. B Maximum-likelihood tree constructed using at least 25 concatenated protein sequences from 469 medium- and high-quality metagenome-assembled genomes (MAGs). Clades are colored according to taxonomic class. From the inner circle to the outer circle, genome characteristics are indicated successively including sample source, circular genomes, N50 of contigs, genome similarity, genome occurrence ratio, and genome completeness. The solid and empty circles correspond to samples from captive and wild giant pandas, respectively. The 40 circular MAGs are indicated by triangles. In the outer layers, genome information (contig N50, genome similarity, and completeness), the occurrence ratio, and average relative abundance in all samples for each MAG are presented, respectively. C Average nucleotide identity (ANI) heatmaps for the 40 circularized MAGs. Taxonomic information is shown on the left. The solid and empty circles correspond to samples of captive and wild giant pandas, respectively. Diamonds represent 17 new species based on a < 95% nucleotide identity threshold. D Rarefaction curves depicting the coverage ratios of sequencing reads in the indicated samples against all sequence reads in all investigated samples. The number of unique identified genes finally reached a saturated state with increasing sample numbers, suggesting complete coverage of the gene catalog. E Percentage of mapping rates (left) and annotation rates (right) of de novo gene predictions generated in this study (as indicated by *) and for previously published giant panda gut metagenome datasets in comparison against the reference catalog pandaGUT