Researchers discover new antibiotic using AI

Researchers at the Massachusetts Institute of Technology in Cambridge, US have discovered a  new antibiotic using artificial intelligence for the first time.

The research team used a machine-learning approach to identify new types of antibiotic from a pool of more than 100 million molecules, including one which reportedly works against a wide range of bacteria.

That included showing efficacy against tuberculosis and strains of bacteria which are considered untreatable or resistant to currently approved antibiotics.

The newly discovered antibiotic – halicin – is the first to be discovered using artificial intelligence, say the researchers.

Despite other research using AI to aid in certain parts of the antibiotic discovery process before now, the researchers – led by synthetic biologist Jim Collins – report that this is the first instance where novel types of antibiotic have been discovered using the AI method.

The team identified both possible investigational candidates and validated promising molecules in animal tests.

The approach can also be applied to other types of drugs, according to Jacob Durrant, a computational biologist from the University of Pittsburgh, Pennsylvania. That could include drugs used to treat cancer or neurodegenerative disease, said Durrant.

The AI algorithm used in the research was modelled after the brain’s structure and designed to learn t properties of molecules atom by atom.

The research team trained the AI’s neural network to identify molecules that inhibit the growth of the common bacteria Escherichia coli (or E.coli) using a collection of 2,335 molecules for which the antibacterial activity was known.

The algorithm learns to predict molecular function without any assumptions about how drugs work and without chemical group being labelled, according to Regina Barzilay, a co-author of the study and AI researcher at MIT.

The researchers used the model to screen a library known as the Drug Repurposing Hub, which contains around 6,000 molecules under investigation for use against human diseases.

The team asked the AI to predict which molecules would work against E.coli and show them molecules which look different to conventional antibiotics.

The research is particularly significant given the growing issue of antimicrobial and antibiotic resistance across the world.

There are approximately 25,000 deaths in Europe alone each year resulting from drug-resistant bacterial infections, with the World Health Organization (WHO) identifying antimicrobial resistance as a global threat to public health.

The study is “a great example of the growing body of work using computational methods to discover and predict properties of potential drugs”, says Bob Murphy, a computational biologist at Carnegie Mellon University in Pittsburgh.

Research and development into new antibiotics has been lacking over the past decade, due in most part to the lack of financial incentive in the area.

GlaxoSmithKline is one of the sole big pharma players still developing new antibiotics, with a new-class antibiotic currently in phase 3 development for use in uncomplicated urinary tract infection (UT) and urogenital gonorrhoea.

Many have called for a new payment model to incentivise the development of new antibiotics, with one proposed plan from the UK government of a subscription-style payment model. That system would pay pharma companies upfront for access to drugs based on their usefulness to the NHS, which has been welcomed by GSK.