Genotypic-phenotypic concordance of Pasteurella multocida isolated from bovine respiratory disease cases.

Pasteurella multocida is a respiratory pathogen that is frequently isolated from cattle suffering from bovine respiratory disease (BRD), a leading cause of morbidity and mortality on modern-day cattle farms. Treatment involves the use of antimicrobials which have been shown to fail in about 30% of BRD cases, in some cases due to antimicrobial resistance. Phenotypic resistance can be confirmed via laboratory antibiotic susceptibility testing (AST) but this requires several days to complete. Genotypic resistance could be quickly assessed via nucleic acid-based assays that target known antibiotic resistance genes (ARGs); however, ARGs associated with antibiotics used to treat BRD, such as tulathromycin, have been shown to have low genotype-phenotype concordance. Hence, this study aims to improve P. multocida genotype-phenotype concordance by applying a machine learning (ML) algorithm to identify novel genomic sequences (biomarkers) with greater accuracy than known ARGs in predicting resistance to antibiotics commonly used to treat BRD. Cultures of P. multocida were isolated from cattle with clinical signs of BRD. Antibiotic susceptibility testing was performed for each isolate. Genomes were sequenced and assembled, followed by the annotation and identification of ARGs using the comprehensive antibiotic resistance database (CARD). A Set Covering Machine algorithm was used to identify novel markers of resistance using the program Kover. ML-generated genomic biomarkers and known ARGs were found to have similar accuracy in predicting the resistance phenotypes for all six antibiotics tested. Antibiotic resistance genes for five of the six antibiotics had ≥ 0.90 accuracy. The ML-generated biomarkers for tulathromycin resistance improved prediction accuracy by 4% compared to known tulathromycin ARGs, the highest improvement by ML compared to ARGs. Interestingly, one ML-generated biomarker for tulathromycin, a macrolide antibiotic, was a mobile element protein upstream of erm(42), mphE, and msrE, three known macrolide ARGs. External validation revealed phenotypic resistance could be accurately predicted using genomic biomarkers determined by ML or ARGs. This study demonstrated that both genomic biomarkers determined by ML and known ARGs can provide an accurate prediction of phenotypic antibiotic resistance in P. multocida isolates. In the future, rapid diagnostic assays could be developed from ML-generated biomarkers or ARG sequences to reduce treatment failures associated with antibiotic-resistant pathogens in cattle suffering from BRD.