Abstract
Additional file 9: Table S4. The mgSs assignment of a sample was dependent upon log10-transformed sequencing coverage in some mgSs (adjusted p < 0.05). For species with multiple significant mgSs (bold), only samples with ≥ 5.5e5 reads were used to train the mgSs classifer. Dependence was tested for using a logistic regression where the response was the mgSs (binary) and the predictor was log10-transformed sequencing coverage. Testing was performed at the participant level. P-values were adjusted using Bonferroni correction.