The very process that created antibiotic resistance – evolution – is now being harnessed to defeat it.
In the shadows of every hospital and ecosystem, a silent war has been raging for millennia. Bacteria and antibiotics have been locked in an evolutionary dance, with drugs developed to kill pathogens and pathogens evolving clever mechanisms to survive them. For decades, humanity has been losing this war, with bacteria evolving resistance to our most potent medicines faster than we can develop new ones. This crisis of antimicrobial resistance, already claiming millions of lives globally, has forced scientists to rethink the very foundations of antibiotic discovery .
Deaths directly attributable to antimicrobial resistance in 2019
But what if we could turn evolution from a foe into an ally? Instead of merely discovering antibiotics that already exist, researchers are now harnessing evolutionary principles to guide the development of new treatments. From resurrecting ancient molecules lost to extinction to accelerating evolution in laboratory settings, a powerful new paradigm is emerging. This approach doesn't just find new antibiotics; it uses the timeless power of evolution to create them, offering a revolutionary strategy to outsmart resistance and safeguard modern medicine for future generations.
Mining ancient genetic material for novel compounds
Accelerating natural selection in laboratory settings
Using algorithms to predict and design new antibiotics
One of the most revolutionary new approaches is molecular de-extinction—the selective resurrection of genes, proteins, or metabolic pathways from long-vanished species. Scientists are essentially treating Earth's evolutionary history as a vast, untapped library of bioactive compounds 1 .
This process relies on two key scientific disciplines:
The potential of this approach was stunningly demonstrated when researchers used deep learning models to discover new antibiotic peptides from extinct organisms. They synthesized 69 predicted peptides, and several showed powerful activity against dangerous bacterial pathogens in mouse models of infection 1 . Our ancestors, and even prehistoric creatures like woolly mammoths and giant sloths, may hold the key to next-generation treatments 4 .
If molecular de-extinction looks backward through time, experimental evolution charges forward. This approach involves placing antibiotic-producing bacteria, such as Actinobacteria, under controlled selective pressure in the laboratory. The goal is to force them to evolve new strategies to kill resistant pathogens, effectively recreating and accelerating the natural optimization process that gave us most existing antibiotics 3 .
In a landmark study, researchers cocultured the bacterium Streptomyces clavuligerus with methicillin-resistant Staphylococcus aureus (MRSA). Over just four months of evolution, the Streptomyces turned on a previously silent genetic pathway and began producing a novel antibiotic that inhibited the growth of the resistant pathogen 3 .
This method has a key advantage: it selects for the complex function of bacterial killing without requiring scientists to understand all the underlying mechanisms. Evolution naturally optimizes for all necessary components, such as membrane penetration and resistance circumvention, in one process 3 .
Artificial intelligence, particularly machine learning, acts as a powerful catalyst that supercharges these evolutionary approaches. By thinking of biology as information code, researchers can devise algorithms to sift through the biological blueprints of both living and extinct organisms to identify sequences with antibiotic potential 4 .
Machine learning models are trained on vast datasets containing the chemical structures of thousands of compounds known to be active or inactive against bacteria.
The trained model then analyzes billions of new or historical chemical structures, predicting which are most likely to have antimicrobial activity.
Generative AI can even design brand-new, "new-to-nature" antibiotic molecules from scratch, exploring a chemical space of nearly infinite possibilities to find promising candidates 4 .
The Collins laboratory at MIT famously used this approach to discover Halicin, a structurally unique compound with potent activity against a broad range of multidrug-resistant pathogens, including Acinetobacter baumannii . This demonstrated that AI could rapidly identify non-obvious candidates that human-driven discovery might never have found.
Chemical compounds screened by AI in hours instead of years
To understand how these principles converge in a real-world experiment, let's examine a pivotal study that resurrected ancient peptides from our extinct relatives.
The experiment, led by researchers at the University of Pennsylvania, followed a clear, step-by-step process 1 4 :
The results were compelling. The algorithms successfully identified several peptides from our extinct relatives that showed potent, broad-spectrum antimicrobial activity.
Furthermore, the researchers discovered that some of these ancient peptides worked even better in pairs, exhibiting powerful synergistic effects.
The scientific importance of this experiment is profound. It proves that evolutionary history is a functionally limitless reservoir for novel drug discovery. These ancient molecules have been "pre-validated" by evolution, having served real biological functions in extinct organisms. They represent structural templates that modern pathogens have never encountered, making them promising candidates for overcoming contemporary resistance mechanisms.
| Peptide Name | Derived From | Skin Abscess Infection Model | Deep Thigh Infection Model |
|---|---|---|---|
| Elephasin-2 | Straight-tusked Elephant | Comparable to Polymyxin B | Comparable to Polymyxin B |
| Mylodonin-2 | Giant Ground Sloth | Comparable to Polymyxin B | Comparable to Polymyxin B |
| Mammuthusin-2 | Woolly Mammoth | Potential activity observed | Not specified |
| Hydrodamin-1 | Ancient Sea Cow | Potential activity observed | Not specified |
| Megalocerin-1 | Giant Elk | Potential activity observed | Not specified |
Source: Adapted from CAS Insights 1
| Peptide Pair | Pathogen Tested | Effect of Combination |
|---|---|---|
| Equusin-1 + Equusin-3 | A. baumannii | MIC decreased by 64 times (reaching sub-micromolar concentrations) |
| Multiple Pairs | A. baumannii & P. aeruginosa | Strong synergy with FIC index as low as 0.38 for A. baumannii |
Source: Adapted from CAS Insights 1 . MIC: Minimum Inhibitory Concentration; FIC: Fractional Inhibitory Concentration. A lower FIC index indicates stronger synergy.
The revolutionary work of evolutionary antibiotic discovery relies on a sophisticated interdisciplinary toolkit.
The raw historical material, sourced from fossilized and subfossil remains, used to reconstruct genetic and protein sequences of extinct organisms.
A high-throughput technology used to read the highly degraded sequences of ancient DNA, enabling reconstruction of fragmented genetic material.
An analytical technique used in paleoproteomics to identify and sequence ancient protein fragments from fossilized remains.
A precise gene-editing tool used to introduce resurrected ancient genes into modern organisms or to engineer improved bacteriophages.
Algorithms that learn from chemical and biological data to predict antibiotic activity in vast digital libraries of molecules.
The chemical components and techniques used to physically synthesize the peptides and molecules identified through computational means.
Despite these challenges, the strategic shift towards evolutionary paradigms represents our best hope. By learning from the deep past, accelerating evolution in the present, and leveraging AI to guide the process, science is building a sustainable engine for antibiotic discovery.
The future of our fight against superbugs may well depend on our ability to partner with the very process that created life's incredible diversity.
Harnessing evolution through molecular de-extinction, experimental evolution, and artificial intelligence offers a revolutionary strategy to outsmart resistance and safeguard modern medicine for future generations.
Resurrect ancient molecules that pathogens have never encountered
Speed up evolution in laboratory settings to generate novel solutions
Use AI to explore chemical spaces beyond human imagination