Abstract
<title>Abstract</title> <p>Rhizosphere microbiome is an important factor in maintaining plant health and development of sustainable agriculture. In this study, bacterial community composition in the canola (Brassica napus) rhizosphere microbiome was analyzed by 16S rRNA amplicon sequencing based on 8,701 samples with 2,663 amplicon sequence variants (ASVs). Analysis of alpha diversity showed that the number of microorganisms per sample equals 61.32 ASVs, while Shannon and Simpson diversity indices are equal to 2.35 and 0.78, correspondingly. Six ASVs were identified as core microbiome components because they were observed in at least 10% of all samples. A microbiome-based health index was developed based on ASV relative abundance and number of ASVs in each sample. A classification model (Random Forest classifier) was built in order to predict two sample groups with different microbiome-based health status. The accuracy of this classifier equals 93.97%, its AUC equals 0.9931, and five-fold cross-validation accuracy equals 93.55 ± 0.59%. Ten ASVs with highest feature importance were identified by feature importance analysis. Thus, the application of machine learning algorithms allowed identifying biologically meaningful patterns in the dataset of the large canola rhizosphere microbiome.</p>