https://bio.nikkeibp.co.jp/atcl/news/p1/26/03/04/14319/

https://www.nature.com/articles/s41467-026-69966-0

Researchers from Kitasato University have developed an AI that has been trained on various ortholog databases covering approximately 2,000 species of bacteria that are publicly available worldwide to optimize codons without changing the amino acids.

By applying the ortholog rules between Ideonella sakaiensis and Bacillus subtiles, a PETase coding gene from Ideonella sakaiensis was converted into a PETase gene with a sequence optimized for Bacillus subtiles. When expressed in B. subtilise, the degradation activity of Bacillus subtilis for PET was more than 10-fold higher, though the PETase gene itself had activity equivalent to or lower than that of the wild-type gene.

According to the researchers, the ortholog-based optimization increased mRNA stability within Bacillus subtilis and favorably affected protein folding activity.

While it is possible to use Ideonella Sakaiensis to decompose PET, Bacillus subtilis has the advantage of forming biofilms, which makes it well suited to decomposing PET film.