Columbia and Harvard scientists reengineered parts of the *E. coli* ribosome to function without the amino acid isoleucine, a move that trims the genetic code from 20 to 19 building blocks. Using AI‑driven protein design, they swapped isoleucine with valine across 36 essential genes and then targeted 21 ribosomal proteins.

The first round of swaps revealed that 17 genes tolerated the change, while 22 proved lethal, and growth slowed in the survivors. Focusing on the ribosome’s small subunit, the team iteratively replaced isoleucine residues, leveraging deep‑learning tools and AlphaFold 2 to predict viable sequences. After testing 16 rplW variants, one design let the organism grow at about 60% of normal speed, surviving without any isoleucine.

These experiments demonstrate that a reduced genetic code can sustain life, challenging the long‑held view that 20 amino acids are indispensable. The work opens pathways for synthetic biology, where streamlined codes could simplify engineered organisms or enable novel biochemistry. The success hinges on AI’s capacity to predict protein folding, marking a milestone for computational biology.