An AI-guided search found PTC, an enzyme attaching glutathione, a sulfur-containing molecule, to bacterial chemicals called phenazines. The coupling was considered spontaneous. The work appeared in PNAS in July 2026.[2]
The March manuscript starts with 14 known enzymes. Related sequences near phenazine-biosynthesis genes expanded training to 593 likely positives, balanced against around 600 negatives. Learning shared sequence patterns then helped rank candidates.[1]
An accelerator for existing chemistry
A catalyst changes how quickly a reaction proceeds without changing its overall thermodynamic balance. That distinction, formalised in IUPAC’s definition, explains why a reaction already known to occur can still conceal an undiscovered enzyme.[3]
After 90 minutes with a purified Pantoea candidate and nickel, pyocyanin was nearly depleted; mass spectrometry detected its glutathione conjugate. Enzyme-free controls remained blue. The enzyme’s physiological role remains unsettled.[1]
A useful follow-up separates chemical capability from biological purpose. Showing that a purified protein helps a reaction tells an experimenter what to investigate inside a cell. Establishing when that assistance matters requires its own evidence.
Sources & further reading
- Discovery of a phenazine–thiol conjugase from sparse data using genome-informed machine learning Shan and colleagues / bioRxiv · 6 Mar 2026 · Accessed 20 Oct 2026
- Published PNAS record PNAS · 27 Jul 2026 · Accessed 20 Oct 2026
- Catalyst definition IUPAC Gold Book · Accessed 20 Oct 2026