Identification of Operons in clostridium difficile

Abstract

Clostridium difficile (C. difficile), an anaerobic, spore-forming, Gram-positive bacterium, is a major nosocomial pathogen that causes severe, toxin-mediated diarrhea and pseudomembranous colitis. Previous studies of pathogenic bacteria, including C. difficile, have investigated the operon structures and their roles in drug resistance and pathogenicity. Despite significant progress in understanding the genome and pathogenesis of C. difficile, predicting operon structures that contribute to its adaptability and virulence remains a challenge. Existing computational tools provide only partial insights, often limited by incomplete genome annotations, environmental variability, and the overwhelming complexity of bacterial regulatory networks. Consequently, there is a need for improved computational frameworks that integrate genomic, transcriptomic, and functional data to accurately predict and analyze operons in C. difficile. Recently, COSMO, an operon predictor developed using ML techniques, was employed to investigate the operon structures and their roles in drug resistance and pathogenicity in Mycobacterium tuberculosis. In this study, the COSMO algorithm was applied to a published Clostridium genomic dataset. Using COSMO and published RNA-seq datasets for Clostridium difficile, we identified and annotated operons in wild-type (WT) and vanS-mutant strains in the presence or absence of vancomycin or ramoplanin. This study (1) assessed the performance of the COSMO model operon predictions using a set of 28 experimentally validated operons (EVOs), and (3) evaluated the functional implications of the genomic organization of operons in response to vancomycin and ramoplanin antibiotic exposure. COSMO correctly identified 23 of the 28 EVOs, yielding high precision (88.46%) but relatively low sensitivity (45%), with an overall F1 score of 59.74%. The COSMO operon predictions were validated using 28 EVOs across WT and vanS mutant strains in the presence and absence of the two antibiotics. The COSMO algorithm proved reliable across the six experimental conditions, correctly identified 27 out of 28 EVOs each time, but failed to predict one EVO (cprABCK operon CD630_RS07455-CD630_RS07465). Fewer operons retain their original size in the presence of vancomycin in WT strain, and both expansions and contractions in operon boundaries were observed. Similarly, exposure to ramoplanin showed that 19 of the 28 EVOs retained their gene organization and gene content. The data suggest that, in a WT strain, both vancomycin and ramoplanin disrupt the underlying operon organization during the stress response.