CdMEC-A: An Automated Python Framework for Spatial Mobilome Profiling and Antimicrobial Resistance Risk Stratification in Clostridioides difficile

Technical Summary: CdMEC-A

Problem Statement
Current genomic surveillance for antimicrobial resistance (AMR) is largely limited to binary presence-or-absence profiling of antimicrobial resistance genes (ARGs). While tools like CARD and AMRFinderPlus effectively catalog resistance determinants, they fail to provide the spatial resolution necessary to determine the mobilization potential of these genes. Specifically, existing pipelines do not distinguish between stable, vertically inherited chromosomal resistance and resistance genes embedded within or adjacent to mobile genetic elements (MGEs) such as plasmids, transposons, and insertion sequences. This lack of spatial context hinders the ability to assess the risk of horizontal gene transfer (HGT) across ecological niches, a critical gap in monitoring pathogens like Clostridioides difficile, where MGEs drive resistance dissemination.

Methodology
To address this, the author developed CdMEC-A, an open-source, automated Python framework designed to quantify the physical base-pair distances between ARGs and MGEs. The pipeline operates through four core functional modules:

1. Homology Search and Feature Discovery: The system ingests FASTA assemblies and performs parallel homology searches. ARGs are identified via BLASTX against the Comprehensive Antibiotic Resistance Database (CARD), while MGEs are detected via BLASTN against a curated, species-specific database of C. difficile mobile elements (including Tn916, Tn5397, Tn6194 families, and various insertion sequences).
2. Spatial Mapping and Distance Calculation: The algorithm iterates through identified ARGs on the same contig to compute exact intergenic distances to upstream and downstream MGEs using a 1-based inclusive coordinate system.
3. Proximity Thresholds and Categorization: Using a standardized 10,000 bp (10 kbp) scanning window, the pipeline classifies mobilization potential into three actionable tiers:
   • Critical-Risk: ARGs embedded directly within MGE boundaries (0 bp distance).
   • High-Risk: ARGs located within 10 kbp upstream or downstream of an MGE.
   • Stable: ARGs located >10 kbp from any MGE or in unlinked chromosomal contexts.
4. Reporting and Visualization: The framework generates structured JSON/TSV reports and visualizations, including linear genomic maps and frequency distribution plots.

The pipeline was validated using a multi-tiered approach:

• Reference Validation: Benchmarking against the C. difficile 630 reference genome (AM180355.1) to verify mapping accuracy against expert annotations.
• Cross-Platform Comparison: Comparing results against RGI (v6.08) and NCBI AMRFinderPlus (V3.12) across 25 distinct isolates.
• Database Curation: The MGE reference database was rigorously filtered (≥70% identity, ≥30% coverage) to accommodate truncated or degraded mobile elements while maintaining specificity via pHMM validation against Pfam domains.

Key Results

• Accuracy: In the C. difficile 630 reference genome, CdMEC-A achieved >99.5% feature mapping accuracy, correctly identifying isogenic copies of erm(B) within the Tn5398 transposon. It demonstrated high coordinate concordance with AMRFinderPlus and successfully rescued four "Pipeline Unique" loci (e.g., cfr(B), clbA) that were missed by RGI due to database curation differences.
• Large-Scale Analysis: Applied to 495 complete C. difficile RefSeq genomes, the pipeline identified 3,154 ARG-MGE associations. The analysis revealed that 10.9% of ARGs were embedded directly within MGE boundaries (Critical-Risk), while 83.2% resided in stable, unlinked chromosomal contexts.
• Performance and Scalability: Computational benchmarking demonstrated that CdMEC-A achieves near-perfect linear scaling at the process level. While task-level threading (increasing threads within a single genome process) yielded negligible speedups due to internal overhead, process-level scaling (running multiple genomes in parallel) was highly efficient. On a 16-core system, the pipeline processed 495 genomes in ~60 minutes, achieving a throughput exceeding 11,800 complete genomes per day with 99.32% hardware efficiency.

Significance and Claims
The paper positions CdMEC-A as a necessary evolution in AMR surveillance, shifting the paradigm from static gene inventory generation to predictive, mobilome-aware risk assessment. The author claims the framework provides:

• Standardized Risk Stratification: A reproducible method to translate raw genome assemblies into biologically meaningful risk tiers (Stable, High-Risk, Critical-Risk).
• High-Throughput Utility: The ability to process thousands of genomes daily, removing the bioinformatic bottleneck of manually inspecting assembly contigs for spatial offsets.
• One Health Applicability: Although validated on C. difficile, the pipeline's architecture is host-agnostic, designed to facilitate future integration into cross-reservoir surveillance networks (human, animal, environmental) to track resistance mobility across ecological boundaries.

The study concludes that CdMEC-A offers a robust, open-source foundation for high-throughput spatial mobilome profiling, enabling more accurate cross-sector risk assessments without requiring manual intervention.