Fuente: PubMed "meat" Braz J Microbiol. 2026 May 11;57(1):137. doi: 10.1007/s42770-026-01913-z.ABSTRACTThis study investigated the antimicrobial resistance (AMR) profile of Salmonella enterica (n = 446) of poultry origin (n = 1020) collected from West Bengal, India. Salmonella Typhimurium (n = 328) were the most frequently isolated serovar and found resistant to tetracycline (72%), ampicillin (57%), and nalidixic acid (53%). Logistic regression revealed significantly higher odds of resistance in isolates from meat, liver, and cecum compared to cloacal swabs. Resistance was also more likely in isolates from broilers and from retail sources compared to other hosts and farms. To predict extended-spectrum beta-lactamase (ESBL) producers, three machine learning algorithms-XGBoost, Random Forest, and LightGBM-were applied. XGBoost demonstrated the highest sensitivity and overall discriminatory power for ESBL prediction, while Random Forest offered balanced performance. Sampling site and sample source emerged as key predictors, with isolates from cecal and meat samples in retail markets having strong association with ESBL production. Genotypic analysis identified blaCTX-M-1 (n = 122) as the predominant ESBL gene, along with tetA (92%) and quinolone resistance determinants. These findings underscore public health threats posed by AMR in poultry-associated Salmonella, particularly the risk of transmission of resistant strains. The observed resistance to critically important antimicrobials reinforces concerns regarding the diminishing efficacy of frontline therapeutic agents. This study also highlights the utility of integrating machine learning with conventional analytics to enhance AMR surveillance and guide targeted interventions in poultry production systems.PMID:42113384 | DOI:10.1007/s42770-026-01913-z