Fig. 3
Evaluation of intermediate precision and interlaboratory reproducibility of SOPs for sequencing library construction and DNA extraction. a, b Distribution of pairwise Aitchison distances of replicated measurements associated with different operator+lot combinations and laboratories, for library construction (a) and DNA extraction (b), as evaluated using the DNA and cell mock community, respectively. Violin plots depict the distribution of all pairwise distances and symbols shown individual datapoints. If applicable, protocol identifiers are indicated. The subpanel in b shows distances between three custom DNA extraction protocols (shown as different shapes) and protocol N. c Bar charts of estimated qmCVs attributed to different components of variance. Error bars for the intermediate precision and interlaboratory reproducibility estimates represent one-sided 95% confidence intervals. d Similar to panel b, for fecal samples. The first two subpanels show results for fecal sample S01. The third subpanel shows distances between different samples (that is, feces from different donors, denoted as S01 to S13) based on measurements performed by the central laboratory, with DNA extraction and library construction performed using protocols N and BL, respectively. The fourth subpanel shows distances between replicated measurements performed by four laboratories for samples S01 to S13, with all steps (that is, DNA extraction, library construction and sequencing) performed by the participating laboratories. Data from laboratories for which at least one of the duplicate measurements was considered an outlier are shown as red symbols in the first and fourth subpanels (see Fig. S17). e Relationship between species-wise coefficients of variation (CVs) of measured relative abundances and mean relative abundances across replicated measurements to calculate the LOQ (fitted CV of 40%) under different levels of replication as indicated by the colors. Note that CV values exceeding 100% were set to 100% for visualization purposes only. The gray line represents the fit of a species’ probability of detection (POD) to its mean abundance to estimate the LOD (see Fig. S18). f Bar charts of estimated qmCVs attributed to different components of variance for fecal sample S01. Error bars for the precision and reproducibility estimates are one-sided 95% confidence intervals. For panels c and f, fill colors show the metric variance component based on which the corresponding qmCV values were calculated (see Supplementary Methods)