With help from artificial intelligence, MIT researchers have designed novel antibiotics that may combat two hard-to-treat infections: drug-resistant and multi-drug-resistant (MRSA).
Using generative AI algorithms, the research team designed greater than 36 million possible compounds and computationally screened them for antimicrobial properties. The highest candidates they found are structurally distinct from any existing antibiotics, and so they appear to work by novel mechanisms that disrupt bacterial cell membranes.
This approach allowed the researchers to generate and evaluate theoretical compounds which have never been seen before — a method that they now hope to use to discover and design compounds with activity against other species of bacteria.
“We’re excited concerning the recent possibilities that this project opens up for antibiotics development. Our work shows the ability of AI from a drug design standpoint, and enables us to take advantage of much larger chemical spaces that were previously inaccessible,” says James Collins, the Termeer Professor of Medical Engineering and Science in MIT’s Institute for Medical Engineering and Science (IMES) and Department of Biological Engineering.
Collins is the senior writer of the study, which appears today in . The paper’s lead authors are MIT postdoc Aarti Krishnan, former postdoc Melis Anahtar ’08, and Jacqueline Valeri PhD ’23.
Exploring chemical space
Over the past 45 years, a couple of dozen recent antibiotics have been approved by the FDA, but most of those are variants of existing antibiotics. At the identical time, bacterial resistance to lots of these drugs has been growing. Globally, it’s estimated that drug-resistant bacterial infections cause nearly 5 million deaths per yr.
In hopes of finding recent antibiotics to fight this growing problem, Collins and others at MIT’s Antibiotics-AI Project have harnessed the ability of AI to screen huge libraries of existing chemical compounds. This work has yielded several promising drug candidates, including halicin and abaucin.
To construct on that progress, Collins and his colleagues decided to expand their search into molecules that may’t be present in any chemical libraries. By utilizing AI to generate hypothetically possible molecules that don’t exist or haven’t been discovered, they realized that it needs to be possible to explore a much greater diversity of potential drug compounds.
Of their recent study, the researchers employed two different approaches: First, they directed generative AI algorithms to design molecules based on a particular chemical fragment that showed antimicrobial activity, and second, they let the algorithms freely generate molecules, without having to incorporate a particular fragment.
For the fragment-based approach, the researchers sought to discover molecules that might kill , a Gram-negative bacterium that causes gonorrhea. They began by assembling a library of about 45 million known chemical fragments, consisting of all possible combos of 11 atoms of carbon, nitrogen, oxygen, fluorine, chlorine, and sulfur, together with fragments from Enamine’s REadily AccessibLe (REAL) space.
Then, they screened the library using machine-learning models that Collins’ lab has previously trained to predict antibacterial activity against . This resulted in nearly 4 million fragments. They narrowed down that pool by removing any fragments predicted to be cytotoxic to human cells, displayed chemical liabilities, and were known to be just like existing antibiotics. This left them with about 1 million candidates.
“We desired to eliminate anything that may appear to be an existing antibiotic, to assist address the antimicrobial resistance crisis in a fundamentally different way. By venturing into underexplored areas of chemical space, our goal was to uncover novel mechanisms of motion,” Krishnan says.
Through several rounds of additional experiments and computational evaluation, the researchers identified a fraction they called F1 that appeared to have promising activity against . They used this fragment as the premise for generating additional compounds, using two different generative AI algorithms.
Considered one of those algorithms, generally known as chemically reasonable mutations (CReM), works by starting with a selected molecule containing F1 after which generating recent molecules by adding, replacing, or deleting atoms and chemical groups. The second algorithm, F-VAE (fragment-based variational autoencoder), takes a chemical fragment and builds it into a whole molecule. It does so by learning patterns of how fragments are commonly modified, based on its pretraining on greater than 1 million molecules from the ChEMBL database.
Those two algorithms generated about 7 million candidates containing F1, which the researchers then computationally screened for activity against . This screen yielded about 1,000 compounds, and the researchers chosen 80 of those to see in the event that they might be produced by chemical synthesis vendors. Only two of those might be synthesized, and considered one of them, named NG1, was very effective at killing in a lab dish and in a mouse model of drug-resistant gonorrhea infection.
Additional experiments revealed that NG1 interacts with a protein called LptA, a novel drug goal involved within the synthesis of the bacterial outer membrane. It seems that the drug works by interfering with membrane synthesis, which is fatal to cells.
Unconstrained design
In a second round of studies, the researchers explored the potential of using generative AI to freely design molecules, using Gram-positive bacteria, as their goal.
Again, the researchers used CReM and VAE to generate molecules, but this time with no constraints aside from the final rules of how atoms can join to form chemically plausible molecules. Together, the models generated greater than 29 million compounds. The researchers then applied the identical filters that they did to the candidates, but specializing in , eventually narrowing the pool right down to about 90 compounds.
They were in a position to synthesize and test 22 of those molecules, and 6 of them showed strong antibacterial activity against multi-drug-resistant grown in a lab dish. Additionally they found that the highest candidate, named DN1, was in a position to clear a methicillin-resistant (MRSA) skin infection in a mouse model. These molecules also appear to interfere with bacterial cell membranes, but with broader effects not limited to interaction with one specific protein.
Phare Bio, a nonprofit that can also be a part of the Antibiotics-AI Project, is now working on further modifying NG1 and DN1 to make them suitable for extra testing.
“In a collaboration with Phare Bio, we’re exploring analogs, in addition to working on advancing the perfect candidates preclinically, through medicinal chemistry work,” Collins says. “We’re also enthusiastic about applying the platforms that Aarti and the team have developed toward other bacterial pathogens of interest, notably and .”
The research was funded, partly, by the U.S. Defense Threat Reduction Agency, the National Institutes of Health, the Audacious Project, Flu Lab, the Sea Grape Foundation, Rosamund Zander and Hansjorg Wyss for the Wyss Foundation, and an anonymous donor.
