AI-Redesigned Starting Points and Outcomes Enhance Protein Evolution

• p. 13 Reference list, entry 8

  Badran, A. H. et al. Continuous evolution of Bacillus thuringiensis toxins overcomes insect resistance. Nature 533, 58–63 (2016).

• p. 13 Reference list, entry 21

  Badran, A. H. et al. Continuous evolution of B. thuringiensis toxins overcomes insect resistance. Nature 533, 58–63 (2016).

• p. 2 Introduction, left column

  have required additional engineering to recover sufficient stability to be used effectively21

• p. 1 Introduction, left column

  and proteases2,6–12

Judge's note: Read both entries on the p.13 page image (crops C022-a and C022-b): same authors, same title except that entry 8 spells out 'Bacillus' where entry 21 abbreviates 'B.', same journal, volume 533, pages 58-63 and year 2016. Both numbers are used in the text: crop C022-d shows the superscript list '2,6-12' on p.1 and crop C022-c shows p.2 '...to recover sufficient stability to be used effectively21'. One work therefore occupies two reference numbers. I did not need outside knowledge: the duplication is visible within the printed list.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 3 Results, 'Protease redesign for stability', right column

  enhanced thermal stability (Fig. 1e and Supplementary Table 5)

• p. 3 Results, 'Protease redesign for stability', right column

  Fig. 1f and Supplementary Table 6

• p. 3 Fig. 1 legend, part e

  Tm values, calculated as the temperature at which the slope of the curve is maximal, are provided in Supplementary Table 6.

• p. 3 Fig. 1 legend, part f

  Resulting kinetic parameters for BoNT/E protease variants are provided in Supplementary Table 5.

Judge's note: Verified on the page images: crop C005-a/c (p.3) shows the Fig. 1 legend assigning Tm to Supplementary Table 6 and kinetic parameters to Supplementary Table 5, and in the same crop the body text assigns thermal stability to Supplementary Table 5 and the kcat/Km values to Supplementary Table 6. Crop C005-e (p.5) shows the Fig. 2 legend making the same assignment as Fig. 1 (Tm in Table 6, kinetic parameters in Table 5). The two legends agree with each other and conflict with the body text, so the two pointers in the p.3 Results are interchanged. The supplement itself is not among the files, so I could not open Supplementary Tables 5 and 6; the defect is the internal contradiction between the text and the two legends. I also checked the only other pointer (p.3, right column: 'Fig. 2c,d and Supplementary Tables 5 and 6'), which cites both tables together and does not disambiguate.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 4 Protease redesign for stability

  toxin heavy chain module for mammalian intracellular delivery (Extended Data Fig. 3f–j)

• p. 23 Extended Data Fig. 3 legend

  (k) Intracellular delivery of HiBiT-tagged redesigned

Judge's note: Opened the full p.22 image and the p.23 legend crop. Panels f–j are the BoNT/X structure map, catalytic rate versus yield, melting curves, Michaelis–Menten fit with the kinetics table, and the PACE-host luciferase time course; none is a delivery experiment. Panel k is the HiBiT luminescence bar chart (HiBiT peptide, BoNT/X WT + HCN, X/D1 + HCN, X/D2 + HCN), which is the delivery result the sentence describes. The range should read 3f–k.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 24 Extended Data Fig. 4b, rows 416 and 418, P2 column

  416

• p. 6 Extended Data Fig. 4b legend

  Negatively charged

Judge's note: Re-rendered panel b from the PDF at high magnification. Reading down the P2 column: SNAP25 D is red, 412 R is blue, 413 I is tan (aliphatic), 414 and 415 D are black (unchanged from SNAP25, so uncoloured), 416 R is red, 417 I is tan, 418 R is red, 419 I is tan. The colouring convention is consistent otherwise - only mutated residues are coloured, and by property - which excludes a deliberate recolouring. The ED Fig. 4 legend confirms the substrates are "color-coded by each mutation's biochemical properties". Arginine is positively charged, so 416 and 418 should be blue like 412. Note the P1' column is coloured correctly throughout (D red, K blue, I tan), so the error is confined to those two P2 cells.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 8 Results, 'Progression of genotype frequencies', right column

  with replicate 1 converging almost entirely on redesign-specific genotypes (Fig. 4c)

• p. 8 Fig. 4 legend

  replicates 1 and 3 for WT-evolved and replicates 2 and 3 for the D3-evolved populations

• p. 7 Fig. 4c, key

  D3 rep2

• p. 8 Results, 'Progression of genotype frequencies', right column

  for two replicates of each starting point for the intermediate evolu- tion

Only the authors can say what is right here.

Judge's note: Read the p.8 passage on the page image (crop C011-a): the clause sits inside a sentence about 'both redesign-initiated lagoons', so the replicate must be a D3 lagoon. Crop C011-b shows the Fig. 4 legend text quoted above. On the upright render of p.7 I read the Fig. 4c/d/e key, which lists exactly 'D3 rep2', 'D3 rep3', 'WT/E rep1', 'WT/E rep3' and 'Flow rate'; the stacked-area panels are labelled Replicate 1 and Replicate 3 for the WT-evolved populations and Replicate 2 and Replicate 3 for the D3-evolved populations. The same paragraph uses the numbering consistently elsewhere ('K225L in replicate 1 and a quadruple mutant in replicate 3 (Fig. 4a)', and K225L is indeed in the WT Replicate 1 panel). Fig. 4b does show a D3 replicate 1 final-timepoint distribution, so the authors may have meant that panel or a different replicate number; as printed, the replicate identifier does not match the cited panel.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 30 Extended Data Fig. 8 legend, part b

  the corresponding ataxin-2 target peptide residue in the alignment in Fig. 4b

• p. 12 Fig. 6 legend, part a

  a, Comparison of native, stepping-stone and final target substrate sequences.

• p. 8 Fig. 4 legend, parts a,b

  High- throughput long-read sequencing quantification of WT-evolved (a) and D3-evolved (b) populations

• p. 11 Fig. 6a, alignment table

  Ataxin-2

Judge's note: Opened the cited target. Fig. 6a (p11) is the alignment: three rows - SNAP25, SS1, Ataxin-2 - across positions P15 to P6', with the scissile bond marked, which is exactly the SNAP-25-to-ataxin-2 correspondence the ED Fig. 8b legend invokes. Fig. 4 (p7, read upright) consists only of read-count bar charts, 3D fraction-of-reads surfaces and time-course line plots; part b is the D3-evolved LoopSeq genotype quantification, per the Fig. 4 legend "long-read sequencing quantification of WT-evolved (a) and D3-evolved (b) populations". There is no alignment anywhere in Fig. 4. I also checked ED Fig. 8b itself (p29), which is the mutation grid built on that correspondence, so the legend is pointing outward to the defining alignment and the pointer should read Fig. 6a.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 16 Methods, 'Medium-scale protein expression and purification'

  Bound proteins were eluted in 4 × 1.5 ml fractions using wash buffer + 250 ml imidazole, pH 7.3.

• p. 15 Methods, 'Small-scale protein expression and purification'

  elution buffer (wash buffer + 250 mM imidazole, pH 7.3)

• p. 16 Methods, 'Toxin purification and luciferase complementation assay for toxin delivery'

  250 mM imidazole

• p. 16 Medium-scale protein expression and purification

  wash buffer + 250 ml imidazole, pH 7.3

Judge's note: Confirmed on the page-16 Methods crop: the printed unit is "ml". The same reagent in the same elution buffer is given as "elution buffer (wash buffer + 250 mM imidazole, pH 7.3)" in the small-scale protocol on page 15, and the toxin purification paragraph on page 16 likewise specifies "250 mM imidazole". The adjacent clause in the very same sentence block reads "wash buffer + 50 mM glutathione", so the surrounding recipes are all concentrations. No stock concentration is given anywhere that would make a volume meaningful. Should read 250 mM.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 16 Methods, 'Toxin purification and luciferase complementation assay for toxin delivery'

  concentrated using a 15-ml Amicon Ultra centrifugal filter (100,000 kDa cut-off) at 3,500g

• p. 16 Methods, 'Toxin purification and luciferase complementation assay for toxin delivery'

  a 0.5-ml Amicon Ultra filter (100 kDa cut-off) at 14,000 g

• p. 16 Toxin purification, concentration step

  (100,000 kDa cut-off)

• p. 16

  15-ml Amicon Ultra centrifugal filter (100,000 kDa cut-off)

Judge's note: Both sentences are on the one page-16 crop and read as quoted. 100,000 kDa is 100 MDa, which would retain nothing of a toxin of roughly 150 kDa, and the second concentration step in the same paragraph specifies 100 kDa for the same chemistry. The intended value is evidently 100 kDa, equivalently 100,000 Da, so the slip is between Da and kDa. The device sizes legitimately differ (15 ml against 0.5 ml), but that does not account for the cut-off.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 30 Extended Data Fig. 8 legend (a)

  green, 580 nm fluorescence detection of the GAPDH loading control

• p. 17 PTEN cleavage assay in mammalian cells

  IRDye 680RD-conjugated secondary antibodies

Judge's note: Read the ED Fig. 8a legend on the p30 page image and the blot itself on p29 (anti-FLAG bands in red, GAPDH band in green). I then searched the whole article text for every mention of GAPDH, IRDye and a detection wavelength: the dye is named IRDye 680RD in both Methods sections that run this western blot (the PTEN cleavage assay and the ataxin-2 cleavage assay, p17 and p18), and '580 nm' occurs exactly once in the paper, in this legend. Nothing in the Methods, the legend or the antibody list describes a second, shorter-wavelength GAPDH secondary that would reconcile the two. The dye designation 680RD and the stated 580 nm cannot both describe the same channel, so one of the two is wrong; the evident fix is to make the legend match the 680RD dye.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 18 Methods, 'Molecular dynamics simulations of protease–substrate complexes'

  Long-range electrostatic interactions were calculated using the particle mesh Ewald method65, and all bonds involving hydrogen atoms were constrained using the SHAKE algorithm

• p. 19 Reference list, entry 65

  Ryckaert, J. P., Ciccotti, G. & Berendsen, H. J. C. Numerical integration of the Cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J. Comput. Phys. 23, 327–341 (1977).

• p. 19 Reference list, entry 63

  Li, P., Roberts, B. P., Chakravorty, D. K. & Merz, K. M. Rational design of particle mesh Ewald compatible Lennard-Jones parameters for +2 metal cations in explicit solvent.

• p. 18 MD simulations

  particle mesh Ewald method65

Judge's note: Read the crops of p18 and p19 at full size. The superscript 65 sits immediately after "particle mesh Ewald method" and before the comma, so it attaches to the electrostatics clause, not the following SHAKE clause. Reference 65's printed title is about integrating equations of motion under constraints for n-alkanes - that is the SHAKE paper on its face, no outside knowledge needed. Ref 63 (Li et al., "Rational design of particle mesh Ewald compatible Lennard-Jones parameters for +2 metal cations") is correctly cited two sentences earlier for the Zn2+ 12-6 model, so it is not the intended PME source either. No primary PME reference appears in the list. The evident fix is to move the marker to the SHAKE clause.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 25 Extended Data Fig. 5b, Tm table column headers

  E(413)1 I162S

• p. 25 Extended Data Fig. 5b, Tm table column headers

  E(413)1 F163S

• p. 25 Extended Data Fig. 5b, melting-curve key

  D3(413)1 I162S

• p. 25 Extended Data Fig. 5a, Tm table column headers

  E(413)1 G154E

Judge's note: Re-rendered the panel b curve key and Tm table from the PDF at high magnification. The melting-curve key lists exactly four traces: D3, D3(413)1, D3(413)1 I162S, D3(413)1 F163S. The Tm table directly beneath it has exactly four columns - D3 54.0, D3(413)1 47.0, E(413)1 I162S 56.0, E(413)1 F163S 46.7 - so the columns map one-for-one onto those traces. The panel's genotype grid is headed S162/S163 and shows D3(413)1 carrying I and F at those positions. Panel a shows the E(413)1 background is defined by E154, G220 and K225, not by residues 162/163, so an I162S or F163S reversion in E(413)1 is not a construct this paper describes. The headers should read D3(413)1.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 4 Results, 'Protease redesign for stability', left column

  while retaining or improving catalytic efficiency compared with the corresponding WT enzyme

• p. 22 Extended Data Fig. 3d, kinetic parameters table, kcat/KM row (µM-1s-1), BoNT/F

  0.082

• p. 22 Extended Data Fig. 3i, kinetic parameters table, kcat/KM row (µM-1s-1), BoNT/X

  0.24

• p. 1 Abstract

  generating variants with improved stability and full catalytic efficiency

Only the authors can say what is right here.

Judge's note: Read the full p.22 image for both Extended Data Fig. 3 tables and the p.4 and p.1 crops for the two claims. Both tables are internally consistent (each kcat/KM equals kcat divided by KM), so the values are not transcription slips. The losses run from 1.4-fold (BoNT/X D2) to 2.2-fold (BoNT/F D2); only the BoNT/E designs improved on WT. I rated this S3 rather than S1 because the same p.4 paragraph describes the BoNT/F results more carefully as "similar catalytic efficiencies", so the defect is an overstatement of degree in the summary sentence and the Abstract rather than a wrong reported number. The authors need to decide how to qualify "retaining" and "full catalytic efficiency".

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send