Designed from scratch, 13 of 69 cysteine proteases cut the peptide backbone

The 19% needs its denominator. Thirteen of 69 is 18.8% ^[12], which leaves 56 designs that did not meet the paper's success criterion ^[13]. The abstract never defines that criterion, and it leaves the 3 x 10^7 figure unplaced among the 13 working enzymes. The paper reports rate enhancements of up to that value ^[5], and "up to" describes the best design, not a typical one.

At this hit rate, a lab would express and assay about five designs for each working enzyme ^[14]. Whether the rate holds for a new target sequence, the abstract does not say.

The structural result is the cleaner one. The designed proteases have folds very different from natural proteases, with a TM score below 0.50 ^[6]. In our reading, that rules out lightly edited copies of a natural scaffold. Six crystal structures sit within 1.2 Angstrom Ca RMSD of the design models ^[7]. Each one checks directly whether the method placed atoms where it said it would. The coordinates are public under six PDB accession codes ^[10], and the design code is on GitHub ^[11].

Rate enhancement is a ratio against a background reaction, and that background is extremely slow: amide bonds are stable for hundreds of years in neutral aqueous solution ^[2]. The ratio can be large while the enzyme itself stays slow. The authors say a gap remains to the most active native cysteine proteases and trace it to imperfect active-site preorganization and substrate positioning ^[8]. None of this gives absolute turnover. Absolute turnover sets how much substrate an enzyme processes per unit time in a cell or a reactor.

The mammalian-cell claim has the same limit. The abstract says the designed proteases efficiently cleave their targets in mammalian cells, and the authors write that this opens the door to a wide range of synthetic biology applications ^[9]. It does not disclose the targets, the cell types or the extent of cleavage. The cleavage is described as sequence-dependent ^[3], but a purified-enzyme assay does not measure cutting at unintended sites inside a cell. For a tool meant to cut one chosen sequence, unintended cleavage is the bottleneck a user will hit.

We think the structures are the firm result: they are verifiable, and the coordinates are deposited. The activity numbers deserve a looser hold until absolute rates and the cell data can be read in full. That view is conditional on the paper's methods matching its abstract, which we have not been able to check beyond the abstract.