Antimicrobial resistance (AMR) is a global health crisis that threatens our ability to treat common infections. Traditional antibiotic discovery has relied on labor-intensive trial-and-error methods, which have been unable to keep pace with the escalating threat of antimicrobial resistance. To overcome this, researchers are utilizing artificial intelligence (AI) to reimagine antibiotic discovery.
Visualizing biology as a vast information source that algorithms can systematically interrogate to uncover hidden molecules, Dr. César de la Fuente, presidential associate professor at the University of Pennsylvania, has developed purpose-built AI models, such as APEX, capable of predicting peptide function solely from sequence.
By systematically analyzing the human proteome for the first time, de la Fuente and colleagues identified thousands of novel antimicrobial compounds that may constitute a distinct branch of peptide-based immunity. Building on this, they extended their efforts to ancient biology, resulting in the discovery of therapeutic molecules from organisms such as Neanderthals and the woolly mammoth.
Technology Networks caught up with de la Fuente at the ELRIG Drug Discovery 2025 event to learn more about the fields of AI antibiotics and molecular de-extinction, and how ancient antibiotics could help combat the global AMR crisis.
Blake Forman (BF):
Senior Science Writer
Technology Networks
Blake pens and edits breaking news, articles and features on a broad range of scientific topics with a focus on drug discovery and biopharma. He earned an honors degree in chemistry from the University of Surrey. Blake also holds an MSc in chemistry from the University of Southampton. His research project focused on the synthesis of novel fluorescent dyes often used as chemical/bio-sensors and as photosensitizers in photodynamic therapy.
Can you discuss how you have used AI to discover new antibiotics?
César de la Fuente, PhD (CDLF):
Presidential Associate Professor
University of Pennsylvania
Dr. César de la Fuente is a presidential associate professor at the University of Pennsylvania, where he leads the Machine Biology Group. He completed postdoctoral research at the Massachusetts Institute of Technology (MIT) and earned a PhD from the University of British Columbia (UBC). He is best known for pioneering computational and artificial intelligence approaches to antibiotic discovery, which have drastically accelerated the time needed to identify preclinical candidates, from years to hours.
We’ve developed AI-based methods that can discover and design new molecules with promising antimicrobial properties. Many of these molecules have progressed to the preclinical stage and have been shown to reduce infections in mouse models. Specifically, we’ve developed AI mining methods that enable us to think of biology as a vast array of information that can be systematically searched to identify functional molecules with antibiotic activity. We’ve now mined across the whole tree of life, including eukaryotes, archaea and bacteria. Not only have we looked at existing biology, but also ancient biology. For example, we’ve identified antibiotics in ancient penguins, woolly mammoths, giant sloths and magnolia trees, as well as the ancient human proteome.
Machine learning and AI have dramatically accelerated the pace of discovery. With traditional methods, it can sometimes take six to seven years to come up with preclinical candidates. With AI, we can discover hundreds of thousands of candidates in just a few hours.
BF:
Senior Science Writer
Technology Networks
Blake pens and edits breaking news, articles and features on a broad range of scientific topics with a focus on drug discovery and biopharma. He earned an honors degree in chemistry from the University of Surrey. Blake also holds an MSc in chemistry from the University of Southampton. His research project focused on the synthesis of novel fluorescent dyes often used as chemical/bio-sensors and as photosensitizers in photodynamic therapy.
You’ve described uncovering antimicrobial molecules from sources as surprising as woolly mammoths. What does this “molecular de-extinction” mean for modern drug discovery, and how realistic is it that such compounds could reach the clinic?
CDLF:
Presidential Associate Professor
University of Pennsylvania
Dr. César de la Fuente is a presidential associate professor at the University of Pennsylvania, where he leads the Machine Biology Group. He completed postdoctoral research at the Massachusetts Institute of Technology (MIT) and earned a PhD from the University of British Columbia (UBC). He is best known for pioneering computational and artificial intelligence approaches to antibiotic discovery, which have drastically accelerated the time needed to identify preclinical candidates, from years to hours.
Molecular de-extinction is a conceptual framework for identifying functional biomolecules throughout evolution and learning how the changes and mutations that occurred in those molecules influenced their functions – for example, their ability to kill bacteria, boost immunity or kill cancer cells.
Some of the candidates, such as mammuthusin, which we identified by mining the woolly mammoth genetic code, are promising. Mammuthusin has shown a good safety and efficacy profile in mouse models; however, the development of this molecule still requires substantial investment, effort and time. So, I would say we’re far from being able to deliver these therapies to humans. However, I believe we have shown progress from a scientific discovery perspective.
BF:
Senior Science Writer
Technology Networks
Blake pens and edits breaking news, articles and features on a broad range of scientific topics with a focus on drug discovery and biopharma. He earned an honors degree in chemistry from the University of Surrey. Blake also holds an MSc in chemistry from the University of Southampton. His research project focused on the synthesis of novel fluorescent dyes often used as chemical/bio-sensors and as photosensitizers in photodynamic therapy.
How could AI-discovered antibiotics change the way we tackle resistant infections in hospitals and communities, and what hurdles remain before they can be widely used?
CDLF:
Presidential Associate Professor
University of Pennsylvania
Dr. César de la Fuente is a presidential associate professor at the University of Pennsylvania, where he leads the Machine Biology Group. He completed postdoctoral research at the Massachusetts Institute of Technology (MIT) and earned a PhD from the University of British Columbia (UBC). He is best known for pioneering computational and artificial intelligence approaches to antibiotic discovery, which have drastically accelerated the time needed to identify preclinical candidates, from years to hours.
The global AMR problem is only going to be solved if we all work together. It’s too big a problem for one single lab to solve. We tend to release our findings as open-access databases so that, together with the scientific community, we can hopefully get closer to effectively combating the AMR crisis.
Bacteria have a hard time developing resistance to some of the small peptides that we have created, which is encouraging. We’re also developing other AI models, such as ApexOracle, which we’re still validating in the lab. This is a multimodal model that can predict and generate molecules active against pathogens whose genomes you input. This has the potential to be effective for combating emerging pathogens.
The next steps in bringing these molecules to the clinic will not only depend on academics like me, but also on big pharma, investors and philanthropists.
BF:
Senior Science Writer
Technology Networks
Blake pens and edits breaking news, articles and features on a broad range of scientific topics with a focus on drug discovery and biopharma. He earned an honors degree in chemistry from the University of Surrey. Blake also holds an MSc in chemistry from the University of Southampton. His research project focused on the synthesis of novel fluorescent dyes often used as chemical/bio-sensors and as photosensitizers in photodynamic therapy.
Do you find researchers quite open to using AI for drug discovery purposes, or do you feel like there’s still some hesitancy?
CDLF:
Presidential Associate Professor
University of Pennsylvania
Dr. César de la Fuente is a presidential associate professor at the University of Pennsylvania, where he leads the Machine Biology Group. He completed postdoctoral research at the Massachusetts Institute of Technology (MIT) and earned a PhD from the University of British Columbia (UBC). He is best known for pioneering computational and artificial intelligence approaches to antibiotic discovery, which have drastically accelerated the time needed to identify preclinical candidates, from years to hours.
When we started over a decade ago, people were extremely skeptical that AI systems could be helpful in complex areas of research like biology or antibiotic discovery. In fact, most people thought it was impossible because biology is too chaotic and multidimensional. For the longest time, we were swimming against the current, and now everybody seems to love AI. I believe that AI can be extremely helpful if trained with high-quality data, but if you don’t have good data, it’s very hard for AI systems to be useful.
Continue reading below…
We have yet to see any AI-designed or discovered molecules approved for use. True success will be achieved when we can bring an AI-designed or discovered drug all the way through clinical trials and see it used to save lives and benefit humanity.
BF:
Senior Science Writer
Technology Networks
Blake pens and edits breaking news, articles and features on a broad range of scientific topics with a focus on drug discovery and biopharma. He earned an honors degree in chemistry from the University of Surrey. Blake also holds an MSc in chemistry from the University of Southampton. His research project focused on the synthesis of novel fluorescent dyes often used as chemical/bio-sensors and as photosensitizers in photodynamic therapy.
Looking forward, what do you see as the next trend in antibiotic discovery?
CDLF:
Presidential Associate Professor
University of Pennsylvania
Dr. César de la Fuente is a presidential associate professor at the University of Pennsylvania, where he leads the Machine Biology Group. He completed postdoctoral research at the Massachusetts Institute of Technology (MIT) and earned a PhD from the University of British Columbia (UBC). He is best known for pioneering computational and artificial intelligence approaches to antibiotic discovery, which have drastically accelerated the time needed to identify preclinical candidates, from years to hours.
I think the concept of multi-modality will become increasingly important. In the end, designing a drug is a multi-objective optimization problem. You want a molecule that effectively kills bacteria but spares human cells, that’s stable enough to perform its function before it degrades and that has acceptable safety and pharmacological properties. You’re always trying to optimize multiple parameters at once.
We’ve developed our ApexDuo model (and are working on other AI models) to address this. It can design molecules with dual functions – for example, peptides that penetrate human cells and target intracellular infections. I think this kind of multimodal design is nearly absent in antibiotic discovery today, and I hope our work helps open new avenues for thinking about drugs in a more integrated, multi-objective way. Ultimately, almost every disease scenario you can imagine would benefit from drugs that can do multiple beneficial things at once.