Describing the nasopharyngeal microbiota of covid-19 patients in Uganda using next generation metagenomic sequencing and machine learning approaches

Abstract

Background: While COVID-19 spread globally, the role of respiratory co-infections in patient outcomes and their prevalence has remained an area of exploration. Prior findings suggest a dominance of bacterial co-infections, notably Streptococcus pneumoniae, Klebsiella pneumoniae, and Haemophilus influenzae. The challenge, however, arises when trying to distinguish colonization from true infection in the COVID-19 environment. Due to limited data from Uganda, coupled with inconsistent diagnostic approaches, this research embarked on utilizing metagenomic sequencing coupled with machine learning to comprehensively profile potential respiratory co-infections among Ugandan COVID-19 patients. Methods: Nasopharyngeal swab samples and data from clinically confirmed COVID-19 patients were collected and stored from different centers during the COVID-19 pandemic. SARS-CoV-2 presence was reconfirmed through real-time RT-PCR assays. 16S metagenomic next-generation sequencing was conducted, followed by bioinformatics analysis using tools such as Kraken2 and R Vegan, Phyloseq and DEseq2 packages. Machine learning was used to determine microbes that were associated with disease severity. Results: Results showed varied nasopharyngeal microbial composition between COVID-19 patients and healthy controls, with higher abundance of specific Operational taxonomic units (OTUs) in COVID-19 patients such Prevotella. Pathogenic bacteria such as Haemophilus influenzae, Klebsiella oxytoca, Salmonella enterica, Staphylococcus aureus and Serratia marcescens had an increased presence in COVID-19 disease states, especially severe cases. Enrichment of opportunistic pathogens, such as Enterococcus species, and depletion of beneficial microbes, like Alphaproteobacteria, was observed between mild and severe cases. Machine learning identified age and microbes such as Ruminococcaceae, Bacilli, Enterobacteriales, porphyromonadaceae, and Prevotella copri as predictive of severity. Conclusion: This research paints a detailed picture of the complex interplay between the nasopharyngeal microbiome, COVID-19 and respiratory co-infections in Uganda. Advanced diagnostics, merged with the prowess of machine learning, can guide future interventions and treatment protocols. Given the findings, clinicians are urged to remain vigilant of potential co-infections to enhance patient care outcomes.