Biomarker Identification for Preterm Birth Susceptibility: Vaginal Microbiome Meta-Analysis Using Systems Biology and Machine Learning Approaches

Abstract

Problem: The vaginal microbiome has a substantial role in the occurrence of preterm birth (PTB), which contributes substantially to neonatal mortality worldwide. However, current bioinformatics approaches mostly concentrate on the taxonomic classification and functional profiling of the microbiome, limiting their abilities to elucidate the complex factors that contribute to PTB.

Method of study: A total of 3757 vaginal microbiome 16S rRNA samples were obtained from five publicly available datasets. The samples were divided into two categories based on pregnancy outcome: preterm birth (PTB) (N = 966) and term birth (N = 2791). Additionally, the samples were further categorized based on the participants' race and trimester. The 16S rRNA reads were subjected to taxonomic classification and functional profiling using the Parallel-META 3 software in Ubuntu environment. The obtained abundances were analyzed using an integrated systems biology and machine learning approach to determine the key microbes, pathways, and genes that contribute to PTB. The resulting features were further subjected to statistical analysis to identify the top nine features with the greatest effect sizes.

Results: We identified nine significant features, namely Shuttleworthia, Megasphaera, Sneathia, proximal tubule bicarbonate reclamation pathway, systemic lupus erythematosus pathway, transcription machinery pathway, lepA gene, pepX gene, and rpoD gene. Their abundance variations were observed through the trimesters.

Conclusions: Vaginal infections caused by Shuttleworthia, Megasphaera, and Sneathia and altered small metabolite biosynthesis pathways such as lipopolysaccharide folate and retinal may increase the susceptibility to PTB. The identified organisms, genes, pathways, and their networks may be specifically targeted for the treatment of bacterial infections that increase PTB risk.

Keywords: bioinformatics; machine learning; pregnancy outcomes; preterm birth; reproductive health; systems biology; vaginal microbiome.