nominally significantp < 0.05
Prior to RF model training, a feature selection was performed, and only species that showed nominally significant differential abundance in t ‐test and/or LinDA analysis ( p < 0.05) were used as features for training.
Prior to RF model training, a feature selection was performed, and only species that showed nominally significant differential abundance in t ‐test and/or LinDA analysis ( p < 0.05) were used as features for training.
Identifying bacteriophages as both highly significant and highly predictive species by complementary methods that differentially process the data supports a real phage signal, rather than a stochastic or spurious result [ 25 , 40 ].