AI Discovers New Antibiotics in Extinct Species DNA

At 16, César de la Fuente made a list of the world's biggest unsolved problems, and one crisis stood out: bacteria that laugh at our best medicines. Two decades later, he's fighting back with a weapon those ancient microbes never saw coming.

De la Fuente, now a bioengineer at the University of Pennsylvania, is teaching artificial intelligence to dig through genetic code and find hidden antibiotic weapons. His team has discovered promising infection fighters in places no one thought to look: the DNA of woolly mammoths, giant sloths, and even our extinct relatives like Neanderthals.

The stakes couldn't be higher. Drug-resistant infections now contribute to more than 4 million deaths every year, and that number could hit 8 million by 2050. Common bacteria like E. coli and staph infections are getting harder to treat, and the pipeline for new antibiotics has nearly run dry.

Traditional antibiotic discovery is expensive, slow, and often hits dead ends. Scientists have spent decades digging through soil and water, hoping to stumble upon the next miracle drug. But with an estimated 1060 possible molecular combinations to explore, that's like searching for a specific grain of sand on every beach on Earth.

De la Fuente's approach flips the script. He treats biology as code that AI can read and understand. His team trains machine learning models to recognize patterns in genetic sequences that signal antibiotic properties, then tests those predictions in the lab.

The results sound like science fiction. His "molecular de-extinction" project has resurrected compounds with names like mammuthusin-2 from woolly mammoths and mylodonin-2 from giant sloths. The team has also found candidates hiding in snake, wasp, and spider venom, plus ancient single-celled organisms called archaea.

Over the past few years, this genetic treasure hunt has built a library of more than one million potential antibiotic recipes. Each one represents a shot at stopping infections that our current drugs can't touch.

James Collins, a synthetic biologist at MIT and pioneer in AI drug discovery, calls de la Fuente "marvelously talented" and credits him with pushing the field forward. Collins knows the territory well. His own team used AI to predict halicin, a broad-spectrum antibiotic now in preclinical development.

Why This Inspires

What makes de la Fuente's work so hopeful is the sheer scope of possibility. Life on Earth has been fighting infections for billions of years, and somewhere in that vast history, organisms evolved defenses we haven't discovered yet. By scanning extinct species and overlooked organisms, he's essentially asking evolution itself for help.

The approach also tackles the economics that have killed antibiotic development. Many pharmaceutical companies have given up on these drugs because they don't generate enough profit. AI dramatically reduces the time and cost of discovery, making the hunt viable again.

At 40, de la Fuente has collected awards from major scientific organizations and recognition as a leader in using AI for real-world problems. But he's not done exploring. His team of 16 scientists at Penn's Machine Biology Group continues scanning genetic codes from every corner of life's family tree.

De la Fuente calls antimicrobial resistance "almost impossible" to solve, but he sees opportunity in that word "almost." For millions of people facing infections that don't respond to treatment, that gap between impossible and almost could mean everything.

The ancient past might just hold the key to a healthier future.