A model for the prediction of antimicrobial resistance in Escherichia coli based on a comparative evaluation of fatty acid profiles

Authors

• Randal S. Stahl, USDA/APHIS/WS,National Wildlife Research Center & St. Jude Children's Research HospitalFollow
• Bledar Bisha, University of WyomingFollow
• Sebabrata Mahapatra, Colorado State University & St. Jude Children's Research Hospital
• Jeffrey C. Chandler, USDA/APHIS/WS,National Wildlife Research CenterFollow

Document Type

Article

Date of this Version

12-18-2019

Citation

The U.S. Government Works

Comments

https://doi.org/10.1016/j.diagmicrobio.2019.114966 0732-8893/Published by Elsevier Inc.

Abstract

Antimicrobial resistance is a threat to agricultural production and public health. In this proof-of-concept study, we investigated predicting antimicrobial sensitive/resistant (S/R) phenotypes and host sources of Escherichia coli (n = 128) based on differential fatty acid abundance. Myristic (14:0), pentadecanoic acid (15:0), palmitic (16:0), elaidic (18:19) and steric acid (18:0) were significantly different (α = 0.05) using a two-way ANOVA for predicting nalidixic acid, ciprofloxacin, aztreonam, cefatoxime, and ceftazidime S/R phenotypes. Additionally, analyses of palmitoleic (16:1), palmitic acid (16:0), methyl palmitate (i-17:0), and cis-9,10-methyleneoctadecanoic acid (19:0Δ) showed these markers were significantly different (α = 0.05) between isolates obtained from cattle and raccoons. S/R phenotype prediction for the above antibiotics or host source, based on linear regression models of fatty acid abundance, were made using a replicated-randomized subsampling and modeling approach. This model predicted S/R phenotype with 79% and 81% accuracy for nalidixic acid and ciprofloxacin, respectively. The isolate host source was predicted with 63% accuracy.