Two studies posted to bioRxiv on November 5 show different ways to use computing in enzyme research. One team used generative AI to design proteins for a specific chemical reaction. The other searched microbial DNA for new CRISPR enzymes and developed a way to test which DNA sequences they can recognize. Both studies report laboratory results and are preprints awaiting peer review.
REXzyme designs enzymes for a given reaction
N. Mimbrero-Pelegri and colleagues built REXzyme to take a description of a chemical reaction and propose amino acid sequences for enzymes that could carry it out. The researchers used it to design variants of formate dehydrogenase that could run the enzyme’s usual reaction in reverse, converting carbon dioxide to formate.
The team produced and purified 20 designs. All 20 showed activity in the lab. In the reported assay, the strongest design converted carbon dioxide 2.5 times as fast per unit of enzyme as the reference protein. The researchers changed amino acids away from the catalytic site, where the chemical reaction takes place. Performance at industrial scale remains untested in the reported work.
NanoPAMDA screens Cas9 enzymes from microbial DNA
Cas9 uses a guide RNA to locate a DNA target. To cut, it also typically needs a short DNA sequence next to that target, called a PAM. Each Cas9 enzyme can favor different PAMs, so finding a promising sequence in a genome is only the first step. Researchers still need to test which targets the enzyme can reach.
I. Genel, C. Suren and I. C. Kurt searched more than 660,000 assembled microbial genomes and additional human microbiome data, narrowing the search to 14 candidate Cas9 proteins. Their NanoPAMDA assay uses nanopore sequencing to track which DNA patterns the enzymes cut over time, with multiple candidates screened in parallel. One candidate, E10, showed sequence-specific cutting in a laboratory assay and was active in E. coli. The result is at the enzyme and bacterial testing stage.