Fig 1.
Sample processing including technical replication of DNA extraction and amplicon library preparation.
Fig 2.
Expected OTU richness as a function of sequence depth (rarefaction).
In silico pooled replicates are shown as dashed lines and individual extraction replicates as solid lines. In panel A richness up to the maximum read number for pooled replicates and in B the same rarefaction curves up to maximum read number for individual replicates. Error bars (in B) represent standard error, for pooled samples calculated as described in Heck et al [47].
Fig 3.
Rarefaction comparing expected OTU richness of pooled samples (grey lines) and individual replicates (black lines).
Only replicates with more than 15,000 reads were included from the samples Fine Sand (A; n = 5), Coarse Sand (B; n = 7) and Clay (C, n = 5). Panel D shows expected richness in pooled samples compared to mean expected replicate richness. Error bars represent standard error, for pooled samples calculated as described in Heck et al [47].
Fig 4.
Rarefaction comparing Shannon diversity of pooled samples (grey lines) and individual replicates (black lines).
Expected Shannon diversity as a function of read sequence depth was estimated using repeated random sub-sampling from replicates with >15,000 reads from Fine Sand (A; n = 5), Coarse Sand (B; n = 7) and Clay (C, n = 5). Panel D shows expected richness in pooled samples compared to mean expected replicate richness.
Fig 5.
Effect of pooled extraction replicate number on OTU richness and Shannon diversity.
Mean diversity estimates based on repeated (i = 100) random selection of replicates to pool followed by random sub-sampling of OTUs to a total sequence depth of 15,000 reads per sample (divided equally between selected replicates). Filtering to protist or metazoan OTUs was carried out after sub-sampling. Error bars represent the standard error of values from the bootstrapping procedure described above.
Fig 6.
Effect of sequence depth on intra-sample divergence.
Divergence is expressed as (A) Bray-Curtis and (B) Jaccard dissimilarity between repeated (i = 100) random sub-samples of all replicates with >15,000 reads. Error bars represent bootstrapped standard errors.
Fig 7.
Effect of sequence depth per replicate on mean silhouette score, representing the separability of replicates based on sample of origin, based on repeated sub-sampling of OTUs.
Filtering to protist or metazoan OTUs was carried out after sub-sampling. Error bars represent bootstrapped standard errors.
Fig 8.
Effect of pooled extraction replicate number on silhouette score.
Mean silhouette scores were calculated based on repeated random selection of replicates to pool and sub-sampling of OTUs to a total sequence depth of 15,000 reads per sample (divided equally between selected replicates). Filtering to protist or metazoan OTUs was carried out after sub-sampling. Error bars represent bootstrapped standard errors.