Neonatal gut microbiota stratification and identification of SCFA-associated microbial subgroups using unsupervised...

Area of Science:

• Microbiome Research
• Infant Health
• Metabolic Profiling

Background:

• The neonatal gut microbiome is crucial for infant health, influencing it via short-chain fatty acid (SCFA) production.
• The precise organization of SCFA-producing microbial communities in neonates is not well understood.
• Understanding these communities is vital for early-life health interventions.

Purpose of the Study:

• To classify distinct microbial subgroups within the neonatal gut microbiome associated with SCFA production.
• To characterize the composition and metabolic potential of these SCFA-producing subgroups.
• To evaluate machine learning models for predicting SCFA-associated microbial clusters.

Main Methods:

• Recruited 71 mother-infant pairs, collecting meconium samples for microbial 16S rRNA gene sequencing and SCFA quantification.
• Applied unsupervised clustering algorithms (K-Means, Agglomerative, Spectral, Gaussian Mixture Model) to identify microbial subgroups.
• Utilized random forest and logistic regression with ROC curves to classify SCFA-associated clusters.

Main Results:

• Agglomerative clustering identified distinct SCFA-producing subgroups, with Cluster 1 enriched in *Bacteroides*, *Prevotella*, and *Enterococcus* showing higher SCFAs.
• A third cluster revealed an intermediate metabolic profile, suggesting a functional continuum.
• Random forest models demonstrated high accuracy (up to 92.98%) in classifying SCFA-associated microbial clusters.

Conclusions:

• Unsupervised clustering and machine learning effectively predict SCFA-associated microbial subgroups in neonates.
• This approach provides insights into early-life metabolic health and microbial community organization.
• Future research can leverage these methods for longitudinal tracking and functional genomic integration.