AMR-Diag: A Novel Diagnostic Tool for Sequence Based Prediction of Antimicrobial Resistance

Infections due to antimicrobial resistance (AMR) bacteria threaten modern health care and are responsible for 700,000 and 33,000 deaths/year globally and in Europe, respectively. A core problem responsible for AMR is incorrect prescription and inappropriate use of antibiotics. The challenge for clinicians managing infection is to precisely guide antibiotic prescription at the point of care (PoC). The time needed by current state-of-the-art methods from collecting the patient samples until the information is available on antibiotic susceptibility patterns is, in the best case, 2-4 days. Thus, there is an unmet need for rapid, innovative, affordable, easy-to-use and PoC infection diagnostics in clinical settings. Taking advantage of the advances in whole-genome sequencing, bioinformatics, and machine learning methods, we have developed a decision-making method AMR-Diag to detect a bacterial infection, including its resistance profile. We have shown that with a sequencing-based approach to blood culture diagnostics, it is possible to identify pathogens and specific AMR-encoding genes using raw nanopore sequencing data obtained within 4 hours after a blood culture is flagged as positive by the incubation system. Identification of pathogens was possible after 10 minutes of real-time sequencing, and all predefined AMR-encoding target genes and plasmids were detected within 1 hour. Additionally, we have found that AMR-Diag machine learning tool networks, coupled with nanopore sequencing, can predict bacterial species, resistome, and phenotype as fast as 1?8 h from the sequencing start. This study holds great promise for future applications in clinical microbiology and health care surveillance purposes. In the future, the AMR-Diag tool can be designed for use by doctors and other healthcare professionals, providing the information needed to choose the best treatment. Thereby resulting in the more prudent and appropriate use of antibiotics.

AMR-Diag aims to reduce the time of detection of bacterial infection and AMR from 2-4 days to <24 hours. This is a giant leap to handle and prevent the development of AMR, which can potentially save millions of human and animal lives in the long run. This study holds great promise for future applications in clinical microbiology and health care surveillance purposes. In the future, the AMR-Diag tool can be designed for use by doctors and other healthcare professionals, providing the information needed to choose the best treatment. Thereby resulting in the more prudent and appropriate use of antibiotics. AMR-Diag achievements will target several UN sustainable development goals: Goal 3 - Good Health and Well-Being, Goal 4 - Quality Education, Goal 9 - Industry Innovation & Infrastructure, Goal 12 - Responsible Consumption & Production Goal 17 - Partnerships for the Goals.

The emergence and spread of antimicrobial resistant (AMR) bacteria is defined as a global health problem by WHO. The situation is at its gravest in low- and middle-income countries, where antibiotic consumption is high and largely unregulated. Due to the lack of real-time diagnostics, prescription of the right antimicrobial at the right time is not always achieved. Time required for culture based identification of pathogen and phenotype-based identification of susceptibility to antimicrobials often necessitates unessential use of broad spectrum antimicrobials, which contributes to increase in resistance among pathogens. Accurate and rapid diagnostics that both identify the pathogen and provide drug susceptibility data in real-time would transform patient management and the current AMR crisis. Their application would reach broadly from primary health care centres to tertiary care hospitals, providing immediate guidance for therapeutic intervention thereby resulting in more prudent and appropriate use of antimicrobials. Taking advantage of the advances in whole genome sequencing (WGS), bioinformatics, proteomics and machine learning methods we plan to develop a decision-making tool AMR-Diag, for the detection of bacterial infection, including its resistance profile. The radical improvement comes from adopting Neural Network based learning for solving WGS data analysis problems. We will cover Extended Spectrum Beta-Lactamases (ESBLs) in Gram negative bacteria with focus on Escherichia coli, Klebsiella, and Acinetobacter. The developed method will be validated using characterized clinical isolates and finally will be evaluated on clinical samples for culture and culture-free identification of pathogens with ESBL resistance profiles. This tool will be easy to run and interpret results from, and will work on standard devices. The proposed project collaboration will strengthen already on-going research activities in each involved research group in India and Norway.