Characterization of the salivary microbiome in COVID-19 infection and development of a cpn60 classifier
The human microbiome is thought to play an important role in many diseases. Recently, many studies have been done on the human microbiome and COVID19, caused by SARS-CoV-2, however only a few have been focused on the salivary microbiome, or oral microbiome in general, even though several oral symptoms of COVID19 are widely reported. In this study, we explored whether there is a difference in the taxonomic composition of the salivary microbiomes of people who are SARS-CoV-2 positive and people who are SARS-CoV-2 negative. We collected saliva samples from 138 volunteers (78 SARS-CoV-2 negative and 60 positive) and identified the salivary microbiota by chaperonin-60 (cpn60) amplicon sequencing. Taxonomic identification of the resulting sequencing was done by alignment to our cpn60 reference database using wateredBLAST. We found that the SARS-CoV-2 positive groups and the SARS-CoV-2 negative groups are different in their salivary microbiome composition on a population scale, however, the two groups are not significantly different in species diversity. Hierarchical clustering resulted in the identification of six clusters. Clusters dominated by Porphyromonas sp. or Prevotella melaninogenica contained significantly more SARS-CoV-2 positive samples than negatives, while clusters dominated by Streptococcus sanguinis or Megasphaera micronuciformis contained more SARS-CoV-2 negative samples. The results of the salivary microbiome study offered an opportunity to explore use of a Naïve-Bayesian classifier an alternative method for identification of cpn60 sequences that would overcome some of the limitations of the wateredBLAST alignment method. We trained the RDP classifier on a curated set of cpn60 barcode sequences, and tested its performance on cpn60 data from three human microbiomes: saliva, vagina, and stool. We found that 62%, 74% and 69% of salivary, vaginal and stool sequence variants were classified to the species level with confidence score >0.8, respectively. These confidently classified sequences accounted for 79%, 86% and 92% of the salivary, vaginal and stool microbiomes, respectively. Identification with the classifier also took significantly less time than wateredBLAST alignment. The results of the studies described in this thesis provide a foundation for future studies aimed at understanding the relationship of oral microbiota to SARS-CoV-2 infection and demonstrate that use of a Naïve-Bayesian classifier is a robust method for taxonomic identification of cpn60 sequences generated in these studies and studies of other microbiomes.
COVID-19, salivary microbiome, cpn60, wateredBLAST, RDPclassifier
Master of Science (M.Sc.)
Microbiology and Immunology