Comparison of sequencing depth
To identify the sequencing depth required to accurately characterize
fecal microbiomes in Sable Island horses, we successively rarefied a
16-sample deeply sequenced shotgun metagenomic dataset. Patterns in
Shannon diversity remained indistinguishable from the minimally rarefied
dataset down to depths of 0.4 million read pairs (Figure 2A), and a
statistically significant decrease in Shannon diversity was only
observed at depths below 40,000 read pairs per sample (t =
-1.959, p = 0.05). However, despite a decrease in the accuracy of
Shannon diversity estimates, Shannon diversity remained highly
correlated between the full dataset and data rarefied to only 2,000 read
pairs per sample (R2 = 0.87, t = 10.24,p < 0.01). Similar patterns were observed with respect
to microbiome beta-diversity. Bray-Curtis distances in the minimally
rarefied dataset were highly correlated with those in more shallowly
rarefied datasets (Figure 1B). For example, Pearson correlation
coefficients remained above r = 0.99 (mantel tests, p< 0.01), at depths as low as 0.2 million read pairs per sample
(Figure 2B). Even at 2,000 read pairs per sample, a strong correlation
was observed with the full sequence dataset
(mantel test: r = 0.81,p < 0.01). Similar results were observed with respect
to MetaCyc reaction and pathway profiles (Figure S3).