Machine learning identifies antibiotic resistant bacteria which can spread between animals and humans.

University of Nottingham experts have developed a new software that combines machine learning and DNA sequencing to help them identify where and how resistant bacteria is being spread between humans, animals, or the environment.

The study was published in PLOS Computational Biology, was directed by Dr. Tania Dontorini, School of Veterinary Medicine and Science.

Anthropogenic environments (spaces created by humans), such as areas of intensive livestock farming, are seen as ideal breeding grounds for antimicrobial-resistant bacteria and antimicrobial resistant genes, which are capable of infecting humans and carrying resistance to drugs used in human medicine. This can have major implications on how certain infections and illnesses can be treated.

A team of experts examined a large-scale commercial poultry farm in China and collected 154 samples from the animals, carcasses, workers, and their homes and environments. They isolated Escherichia coli (E. coli) from the samples. Although these bacteria can be quite harmless in the gut of a person, they can also cause severe stomach cramps and diarrhea.

Researchers used a combination of machine learning, whole genome sequencing and gene sharing networks to identify the different types pathogens in the farm. Antimicrobial genes, which confer resistance to antibiotics, were found in both pathogenic as well as non-pathogenic bacteria.

Machine learning enabled the team to discover a network of genes that are associated with antimicrobial resistant genes. This network was shared by animals, farm workers, and the environment. This network included genes known for causing antibiotic resistance, as well as unknown genes associated with antibiotic resistance.

Dr. Dottorini states that while they cannot pinpoint the origin of the bacteria at this point, they can only say that they discovered it and that it has been shared among animals and humans. This is concerning, as people can become resistant to drugs in two ways. They can either directly contact an animal or eat contaminated meat. This could be a problem in poultry farming as it is the most commonly consumed meat in the world.

“The computational tools we have created will allow us to analyze large complex data, from multiple sources, and identify hotspots for specific bacteria. They are fast, precise, and can be applied to large environments such as multiple farms simultaneously.

“We know a lot about antimicrobial resistance genes, but how can we get beyond them and uncover new targets to create new drugs?”