Summary
A bioRxiv preprint describes four rounds of computational screening that identified natural hydroxymethylfurfural oxidases with improved activity on a key FDCA-production reaction. Some candidates approached an engineered benchmark under the tested conditions.
A bioRxiv preprint posted on 5 October 2026 reports the discovery of natural hydroxymethylfurfural oxidases (HMFOs) with improved activity in a key step toward producing 2,5-furandicarboxylic acid (FDCA). The authors describe a search conducted over four rounds, combining molecular simulations, experimental results and machine learning to refine which enzymes to investigate. Some of the natural candidates approached the performance of the engineered benchmark 8BxHMFO under the conditions tested.
Why the FFCA reaction matters
The process examined converts 5-hydroxymethylfurfural (HMF) into FDCA, a chemical used to make the polyester polyethylene furanoate (PEF). The final oxidation converts 5-formyl-2-furancarboxylic acid (FFCA) into FDCA. The authors identify this reaction as a principal bottleneck for industrial application.
HMFOs are enzymes that can catalyse this oxidation. Finding naturally occurring versions with stronger activity on FFCA could provide promising biocatalysts for this part of the production route. The preprint focuses on identifying such enzymes, rather than reporting an industrial production process.
A search refined over four rounds
The computational campaign began with conservative screening based on sequence homology—the similarity between an enzyme’s amino-acid sequence and known sequences. In later rounds, the researchers expanded to large-scale sequence mining assisted by machine learning. They say that each iteration was redesigned in response to limitations identified through molecular simulations and experimental characterisation.
The authors report several natural HMFOs with substantially improved FFCA-oxidising activity. Some performed close to 8BxHMFO, an engineered enzyme used as a benchmark. The abstract describes this comparison qualitatively; it does not provide activity values or assay conditions, so the size of the performance gap cannot be assessed from the available text.
The work illustrates an iterative approach to enzyme discovery: computational predictions guide experimental testing, and the resulting evidence informs the next search. Its reported candidates may help broaden the set of enzymes available for further development of the FDCA production route. The findings are presented in a preprint.