AI Designs DNA Faster Than It Can Be Manufactured: NunaBio Identifies a New Limit in Synthetic Biology

NunaBio assembles complex DNA sequences outside cells when repeats or bacterial toxicity cause conventional synthesis to fail

On August 21, NunaBio CEO Joe Hedley described the gap between a sequence on a computer and a laboratory sample in an interview with SynBioBeta. The company calls its method NunaSynth. It assembles DNA outside living cells from components whose properties have been tested in advance.

Computational models give researchers new DNA sequence designs. The experiment begins once the laboratory receives a physical sample and can test it.

In conventional synthesis, short DNA fragments are first produced chemically and then joined into a longer construct. Many methods also amplify the construct in bacteria. Repeats, regions with a particularly high proportion of G and C bases, and long stretches of A bases can interfere with assembly. Some constructs are toxic to bacteria. Failure at any of these stages leaves the researcher without a sample to test.

In a recent study of AI-designed bacteriophage genomes, involving viruses that infect bacteria, the authors had to synthesize and assemble 285 candidates after computational selection. Only then could they test the candidates in bacterial culture.

NunaBio describes a different method. The complete construct is assembled outside living cells from components whose properties have been tested in advance, without amplification in bacteria. The company calls this method NunaSynth.

Hedley proposes evaluating manufacturing by asking a different question:

“The price per DNA base measures the supplier’s economics, not whether the customer’s research can be carried out reliably at all.”

Price captures only part of the problem. A physical sample starts the next step in the cycle. Researchers test it and use the result to select the next sequence. The ability to manufacture DNA therefore determines which designed variants reach the experimental stage.