Data Challenge: AMR-Vision: MKULAB’s AI-Powered Laboratory Surveillance Tools

Surveillance of Acinetobacter baumannii–calcoaceticus complex clinical isolates collected between 2016 and 2021 is critical for enhancing patient health, strengthening antibiotic stewardship, guiding public health, and fortifying health systems. This bacterium is notorious for causing severe hospital-acquired infections, especially in intensive care settings, with high morbidity, mortality, and treatment costs. By integrating whole-genome sequencing data (WGS), epidemiological metadata, and machine learning–augmented resistance profiling. Assemblies will undergo quality control (FastQC, Trimmomatic, SPAdes/Unicycler). We will infer phylogenetic relationships using core-genome single-nucleotide polymorphisms (SNPs) and multilocus sequence typing (MLST). Resistance gene detection will leverage our in-house AI/ML pipelines—BacARscan, BacEffluxPred, and β‑LacFamPred—capable of detecting ARGs, efflux pumps, and β‑lactamase genes even in fragmented reads or contigs. BacARscan, based on HMM and machine-learning models, demonstrates ~92 % precision and 95 % F‑measure for ARG detection directly from reads. BacEffluxPred, an SVM-based two-tier model, classifies efflux proteins with Tier-I accuracy of ~86 %, and family-level accuracies between 92 %–99 %. β‑LacFamPred utilizes hidden Markov models to predict Ambler-class β‑lactamases and 96 families with near-perfect validation performance. We will integrate ARG, efflux, and β‑lactamase profiles with phylogenies, phenotypic susceptibility data, and patient metadata to map temporal and spatial patterns of resistance emergence and transmission clusters. Spatio-temporal modeling (e.g., SaTScan, R) will identify hotspots and guide targeted infection-control interventions. ARG data will be fed into regional and global surveillance platforms, such as EARS-Net and the Global Microbial Identifier framework. The proposed methodology combines cutting-edge bioinformatic genomics analyses, artificial intelligence, and machine learning models to deliver actionable insights, enabling clinicians to choose effective therapies more quickly, preserve antibiotic potency, inform stewardship, and empower health systems with data-driven tools to prevent and respond to AMR threats. In sum, this research, anchored by WGS, AI-powered detection, and epidemiological modeling, will empower clinicians with precise resistance information, reinforce stewardship practices, inform public health strategies, and build resilient health system capacity to respond to antimicrobial resistance.