Scientists use AI to create bacteria that run key cellular machinery on just 19 amino acids

AI Transforms the Fundamental Alphabet of Life

All life on Earth operates using the same molecular vocabulary: 20 amino acids that cells assemble into proteins. But researchers have now achieved what seemed impossible just years ago—reengineered E. coli bacteria to run essential cellular machinery with fewer than 20 amino acids 1

1

. The work, published in Science, demonstrates how AI-guided protein design can rewrite biology's fundamental rules while offering clues about how early life forms might have functioned with simpler chemistry.

Harris Wang, a synthetic biologist at Columbia University, led the effort to subtract isoleucine from the bacterial proteome 2

2

. His initial attempts using simple substitutions failed—fewer than half of the modified proteins remained functional. The project languished until artificial intelligence tools transformed what was possible. Systems like AlphaFold can predict protein structures, while protein language models such as ESM2 and MSA Transformer suggest entirely new amino-acid sequences that fold and function properly 3

3

.

Targeting the Ribosome: Biology's Final Boss

Rather than tackle all 4,000-plus proteins in Escherichia coli, Wang chose an ambitious target: the ribosome. This complex of more than 50 proteins sits at the heart of cellular machinery, translating genetic instructions into proteins 1

1

. "Like in a video game, we just pushed the 'skip to the final boss' button," Wang explained 3

3

.

The team selected isoleucine for removal because analysis showed it was frequently substituted with structurally similar amino acids like valine and leucine across bacterial species 2

2

. Their first brute-force approach—simply swapping isoleucine with valine or leucine—worked for 18 out of 50 ribosomal proteins, but the resulting bacteria grew poorly, achieving only 40 percent fitness compared to normal cells 3

3

.

When AI Reveals Non-Intuitive Solutions

Wang partnered with computational biologists Sergey Ovchinnikov and Simon Kozlov at the Massachusetts Institute of Technology to leverage AI models informed by evolution and protein structure prediction 1

1

. The protein language models proposed evolutionarily plausible mutations that simple swaps would miss, while structure-based tools like AlphaFold2 and ProteinMPNN verified the redesigned proteins would fold correctly.

The AI proposals surprised even the researchers. While redesigning a ribosomal protein called RpsJ, the system remodeled an alpha helix and introduced eight new nearby mutations to compensate for substituting just two isoleucines 3

3

. "Some of these AI designs were really surprising," Wang noted. "They didn't look like anything we would have anticipated."

Yet AI alone couldn't solve everything. A few proteins required hands-on laboratory work, highlighting both the power and limits of current technology. "We have not solved biology yet with AI," Kozlov acknowledged. But compared to previous capabilities, "this is a dramatic acceleration" 1

1

.

Engineering Cells That Defy Nature's Standard

The team successfully replaced all 382 isoleucines across the ribosome's protein components. Their final strain, Ec19, carries 21 isoleucine-free ribosomal proteins out of 52 total 3

3

. The reengineered E. coli bacteria grew robustly with only minor slowdown, maintaining above 90 percent fitness compared to unmodified cells, and remained genetically stable for more than 450 generations 1

1

.

"It's a tremendous tour de force," says Kaihang Wang, a synthetic biologist at the California Institute of Technology. However, he adds, "it's a first baby step of a grand journey" toward building an entire cell running on a 19-amino acid alphabet 1

1

. Reducing the amino acid alphabet entirely would push the limits of life's chemistry and enable new kinds of synthetic organisms.

Implications for Medicine, Biotechnology, and Origins of Life

The research offers practical pathways for protein engineering beyond natural constraints. Removing an amino acid "frees up" the DNA sequences that typically code for it, allowing those sequences to be reassigned to encode synthetic amino acids for creating new drugs or molecules 2

2

. Tom Ellis, a researcher in synthetic genomes engineering at Imperial College London, notes the findings showcase AI's ability to predict protein structures and could simplify creating designer proteins for medicine and biotechnology 2

2

.

The stability of the modifications over 450 generations suggests something profound about biology's past. "That finding lends support to the idea that [early] life was probably just fine for a while with a smaller palette," says Christopher Snow, a protein engineer at Colorado State University 2

2

. The work hints that organisms predating even Earth's common ancestor may have operated on leaner chemistry.

Wang's team plans to apply this approach across the rest of the genome engineering toolkit and perhaps attempt creating bacteria with just 18 amino acids. Organisms with reduced dependence on particular amino acids might better survive hostile environments or resist viral infections 2

2

. But as Wang notes in an accompanying commentary, reaching these milestones will require combining advanced computational design with human ingenuity. "The AI is tremendously powerful," he says. "But human input is still critical—at least for now" 1

1

.