Bioinformatics Approaches to the Understanding of Molecular Mechanisms in Antimicrobial Resistance.

Pieter-Jan Van Camp,Aleksey Porollo,David B Haslam

doi:10.3390/ijms21041363

Abstract

Antimicrobial resistance (AMR) is a major health concern worldwide. A better understanding of the underlying molecular mechanisms is needed. Advances in whole genome sequencing and other high-throughput unbiased instrumental technologies to study the molecular pathogenicity of infectious diseases enable the accumulation of large amounts of data that are amenable to bioinformatic analysis and the discovery of new signatures of AMR. In this work, we review representative methods published in the past five years to define major approaches developed to-date in the understanding of AMR mechanisms. Advantages and limitations for applications of these methods in clinical laboratory testing and basic research are discussed.

Highlights

Antimicrobial resistance (AMR) contributes to antibiotic (AB) treatment failure and increasing rates of mortality in various infectious diseases
We summarize the current state of bioinformatics methods for the discovery of AMR molecular mechanisms and discuss advantages and limitations of machine leaning (ML) models in the prediction of AMR and their use in clinical settings
Bioinformatics approaches to the AMR research can be broadly categorized as those that focus on fast and reliable predictions of AMR to be applied in clinical settings, and those that explore the underlying molecular mechanisms of AMR

Summary

Introduction

Antimicrobial resistance (AMR) contributes to antibiotic (AB) treatment failure and increasing rates of mortality in various infectious diseases. Not all molecular mechanisms underlying AMR are uncovered or understood yet, especially in Gram-negative bacteria This means that currently the most reliable way for evaluating resistance to AB in clinical microbiology laboratories is to grow organisms in culture, expose them to various antibiotic concentrations, and assess the impact on growth. With the advent of whole genome sequencing (WGS) permitting quick access to de novo sequenced genomes of pathogens and the collection of sufficiently large datasets of clinical isolates, it is plausible to employ advanced bioinformatics approaches (e.g., machine learning) to gain new insights in the more complex molecular mechanisms of AMR These approaches can evaluate many samples at once by exploring different types of data (e.g., genomics- or metabolomics-based) and reveal new insights previously not attainable. Sci. 2020, 21, 1363 the chronological order of their original publication, but will subsequently be grouped and discussed per approach they employ

Approach 1

Approach 2

Approach 3

Approach 4

Conclusions