Identification and Characterization of Microbial Signatures and Potential Biomarkers for Irritable Bowel Syndrome in Bangladesh

Abstract

Introduction: Irritable Bowel Syndrome (IBS) is a functional gastrointestinal disorder increasingly linked to gut microbiome dysbiosis. However, metagenomic studies integrating taxonomic, functional, resistome, virulome, and genome-resolved analyses in Bangladeshi IBS patients are limited. This study applied whole-metagenome sequencing (WMS), comparative genomics, and machine-learning (ML) approaches to characterize gut microbiome alterations in IBS patients relative to healthy controls (HC). Materials and Methods: A total of thirty (n=30) stool samples were collected from IBS-diagnosed patients (n=20) and HC (n=10) at the National Gastroliver Institute & Hospital, Mohakhali, Dhaka, Bangladesh. Of these, based on the quality and concentration of extracted total genomic DNA, ten IBS samples (n=10) and six HC samples (n=6) were selected for WMS. Sequencing was performed using the Illumina Next Generation Sequencing (MiSeq 4000) platform. The raw FASTQ sequences were subjected to quality control using the Trimmomatic tool. Downstream analyses were performed using two complementary pipelines: the Chan Zuckerberg ID (CZID, formerly IDseq) for host filtering, taxonomic profiling, diversity metrics, and assembly, as well as the Kraken2-based workflow for antimicrobial resistance (AMR) and virulence factor genes detection. Metagenome Assembled Genomes (MAGs) were reconstructed using MEGAHIT, MetaBAT2, and MaxBin2. Comparative global analysis involved a total of 1,109 samples (HC=492; IBS=617), comprising 1,093 publicly available gut metagenomes from across all continents, along with the Bangladeshi samples (n=16). ML models were trained on combined global and local taxonomic datasets after normalization, feature selection, and hyperparameter optimization. Results: IBS patients exhibited clear dysbiosis in their gut microbiome characterized by a marked reduction of Actinobacteria, especially Bifidobacterium spp., and increased abundance of Firmicutes and pathobionts such as Klebsiella pneumoniae, Enterobacter cloacae, and Sutterella wadsworthensis. Notably, multiple Bifidobacterium species, including B. breve, B. bifidum, B. catenulatum, B. pseudocatenulatum and B. longum subspecies, were significantly enriched in the HC samples compared to IBS patients. AMR profiling revealed a relatively higher prevalence of tetracycline, macrolide, and β-lactam resistance genes in IBS samples. MAG analysis recovered 392 genomes, with IBS-derived MAGs harboring more AMR and virulence genes. Comparative Page | ii analysis of global datasets, including Bangladeshi samples, revealed significantly lower alpha diversity in IBS patients supported with Chao1 (p-value: 2.6226e-05), ACE (p-value: 1.8209e-05), and Observed (p-value: 3.0546e-05) indices, indicating higher species richness in HC. Beta diversity analysis showed clear and distinct clustering of IBS and HC samples, reflecting significant differences in overall community composition. LEfSe analysis (LDA score > 2.5, p < 0.05, FDR-adjusted) identified distinct microbial signatures between IBS patients and HC groups, with healthy individuals enriched in Bacteroides spp. (B. stercoris, B. uniformis) and Bifidobacterium catenulatum. In contrast, IBS patients showed increased abundances of Klebsiella, Roseburia, Faecalibacterium, and uniquely associated taxa including Collinsella aerofaciens, Klebsiella pneumoniae, Streptococcus salivarius, and Clostridium spp. Among the developed ML models, Logistic Regression demonstrated superior IBS prediction performance with an accuracy of 0.80 (ROC-AUC 0.88). Conclusion: This multilayered metagenomic investigation revealed substantial microbial, functional, and genomic disruptions in Bangladeshi IBS patients, consistent with patterns observed in global datasets. The findings underscore the potential of microbiome-based ML models in early diagnosis and targeted therapeutics for IBS. Nevertheless, further studies with larger, multi-center, and longitudinal cohorts are warranted to validate these findings and resolve population- and subtype-specific microbiome signatures in IBS.