Researchers have discovered nearly one million potential antibiotic compounds hidden within the vast genetic data of microbes, a breakthrough that could revolutionize the fight against superbugs. This achievement, powered by artificial intelligence (AI), marks one of the largest antibiotic discovery efforts to date and offers immense hope for the future of medicine.

In a groundbreaking discovery, an international team of scientists has identified nearly one million potential antibiotics, promising new hope in the fight against antimicrobial resistance (AMR). Using machine learning, the researchers analyzed vast genomic datasets to pinpoint these novel antimicrobial peptides (AMPs), marking a significant leap in antibiotic discovery.

The Power of AI in Antibiotic Discovery

Discovering a new antibiotic using conventional methods typically takes about 10 to 15 years. This lengthy process begins with identifying potential compounds and screening them for antimicrobial activity. Once a promising candidate is found, it undergoes preclinical testing in laboratories and on animals to assess safety and efficacy. If successful, the antibiotic then enters multiple phases of clinical trials involving human volunteers to ensure it is both effective and safe for public use. Throughout this time, researchers work to understand its mechanism of action, potential side effects, and optimal dosages. Only after passing these stringent tests and regulatory reviews can the new antibiotic finally reach the market.

While there’s not much you can do about shortening clinical trials, AI has proven to be a fantastic tool for drug discovery. These computer algorithms can rapidly analyze vast datasets of chemical compounds, identifying potential candidates in a fraction of the time it takes traditional methods.

For their new study, researchers at the Queensland University of Technology and the University of Pennsylvania Utilizing employed machine learning to analyze the genomes of tens of thousands of bacteria and other microorganisms. This analysis revealed a treasure trove of 863,498 candidate antimicrobial peptides, molecules with the potential to kill or inhibit harmful microbes.

“We think this is the largest exploration ever described of biological data as a source of antibiotics. It’s incredibly exciting because it opens up a vast new resource for potential antibiotics, significantly expanding our toolkit against resistant bacteria. We were pleasantly surprised by the sheer volume and diversity of potential antibiotics we uncovered. This shows just how much potential there is in unexplored biological information like microbial data,” co-author Cesar de la Fuente-Nunez of the University of Pennsylvania told ZME Science.

A New Era in Combatting Superbugs

Over 90% of these peptides were previously unknown. So, to validate these findings, the researchers synthesized 100 peptides and tested them against 11 medically significant bacterial strains in the lab, including antibiotic-resistant varieties like E. coli and Staphylococcus aureus. Remarkably, 63 of these peptides eradicated the growth of at least one pathogen, with some demonstrating effectiveness at very low doses.

Impressively, the peptides showed promise even in preclinical models. Some reduced bacterial loads in mice by up to four orders of magnitude, comparable to the effects of the antibiotic polymyxin B. These findings indicate that these newly discovered peptides could become vital tools in our antibiotic arsenal.

“A memorable moment during our research was when we first saw that the compounds discovered were active in ground-truth experiments. It was a moment of validation for all the hard work and felt like we had just scratched the surface of something monumental,” de la Fuente-Nunez said.

The findings also shed light on how these peptides work. Many appear to target bacteria by disrupting their outer membranes, essentially bursting the bacterial cell like a balloon.

Unlocking the World’s Microbial Dark Matter

The rise of superbugs poses a severe threat to global health, with antimicrobial resistance (AMR) causing approximately 1.27 million deaths annually. Projections suggest that without new antibiotics, this number could escalate to 10 million deaths per year by 2050. This urgent need for novel antibiotics has driven researchers to explore innovative methods like AI to discover effective antimicrobial compounds.

Nature has always been a prolific source of antibiotics. Bacteria have evolved numerous defenses against other microbes, often in the form of short proteins called peptides. The AI-driven approach of this study allowed researchers to tap into the “microbial dark matter” — genetic information from a wide array of environments including human and animal guts, soil, marine environments, and more.

In contrast, the discovery of penicillin by Alexander Fleming in 1928 was a serendipitous event rather than a targeted search. Fleming noticed that a mold, Penicillium notatum, had contaminated one of his bacterial culture plates and killed the surrounding bacteria. This accidental observation led to the isolation and identification of penicillin, the first true antibiotic.

The traditional discovery of penicillin relied on natural observation and painstaking manual processes. Now, the modern AI-driven approach uses computational power to explore microbial diversity at an unprecedented scale. AI can rapidly analyze genetic information from diverse environments, predicting and pinpointing new antibiotic candidates that human researchers might miss.

Moving the discovery forward

“The next steps involve rigorous testing in various preclinical models to ensure safety and efficacy. After that, we would need to conduct Phase I clinical trials in humans, which focus on safety. If those are successful, we would proceed to Phase II and III trials to further test efficacy and safety in larger human populations. Optimistically, it could take around 5-7 years before these new antibiotics might be available for human use, depending on the results of each phase of testing,” de la Fuente-Nunez said.

The research team has made their repository, named AMPSphere, publicly available. This open-access database could be a game-changer, providing a wealth of potential leads for antibiotic developers. The hope is that by leveraging this data, the pace of antibiotic discovery will quicken, bringing us closer to effective solutions against resistant bacteria.

“We have made all these data, information, and code freely available for anyone to access, with the hope of advancing science and benefiting humanity. By making all our data, findings, and code freely available to the scientific community, we aim to accelerate the discovery and development of new antibiotics. This collaborative approach can help scientists worldwide to build on our work, share insights, and make faster progress towards solving the antibiotic resistance crisis. Collaboration and open science are key to tackling such a global health challenge,” de la Fuente-Nunez said.

The findings appeared in the journal Cell.