Fig. 7: Quality of ulrb clustering measured by the average Silhouette score as a function of number of samples, ASVs, and sequencing depth.
From: Definition of the microbial rare biosphere through unsupervised machine learning
a Number of samples (n), ranging from 6 samples up to 114 samples, in increments of two samples; b number of ASVs, ranging from 100 to 4000 ASVs per sample, in increments of 300 ASVs; c, number of reads per sample, in increments of 1000 reads. The samples (a), ASVs (b) and reads (c) were randomly selected in each increment, without replacement. For the response variable, the mean (±sd) of the average Silhouette score was used and the classifications were grouped by different colors, as illustrated in the figure legend. The MOSJ2016-2020 dataset was used for this figure.
