We appreciate the opportunity to comment on Dr. Salunke’s response. Regrettably, we find the response unsatisfactory, as he and his colleagues suggest that it is acceptable to project arbitrary structural models into electron density noise.

All of the seven structure models in question are plagued by the same issues, so we will illustrate this with only one example. Figure 2E in Dr. Salunke’s response (showing the difference density map for the 4H0H model) shows a choppy pattern of electron density contoured at 2.5 σ level that to some extent covers the proposed peptide model.

FIGURE 2.

Positive omit electron density (green) and negative omit electron density (red) of the Ab fragment in PDB entry 4H0H, generated by CNS. 2.5 σ level for both contours. The tyrosine (Y240) in the center was omitted from the model.

FIGURE 2.

Positive omit electron density (green) and negative omit electron density (red) of the Ab fragment in PDB entry 4H0H, generated by CNS. 2.5 σ level for both contours. The tyrosine (Y240) in the center was omitted from the model.

Close modal

We have repeated map calculations using identical software (the Crystallography and NMR system [CNS] 1.3). The positive difference density presented by Dr. Salunke matches our calculations. However, a difference density map is universally presented in publications with both positive (green) and negative (red) contoured levels. The image shown in Fig. 1 reveals why this is important.

FIGURE 1.

Positive omit electron density (green) and negative omit electron density (red) for chain P in PDB entry 4H0H generated by CNS. 2.5 σ level for both contour levels. This figure and Figs. 24 were rendered using Coot (1).

FIGURE 1.

Positive omit electron density (green) and negative omit electron density (red) for chain P in PDB entry 4H0H generated by CNS. 2.5 σ level for both contour levels. This figure and Figs. 24 were rendered using Coot (1).

Close modal
FIGURE 4.

Positive omit electron density (green) and negative omit electron density (red, none visible at this level) for chain P in PDB entry 4H0H generated by CNS. 6 σ level for both contours.

FIGURE 4.

Positive omit electron density (green) and negative omit electron density (red, none visible at this level) for chain P in PDB entry 4H0H generated by CNS. 6 σ level for both contours.

Close modal

Whereas some positive electron density is present at the 2.5 σ contouring level, so is an equal amount of negative density. If the claim is that the positive density at this level reflects the physical presence of the peptide molecule, then what is the meaning of the negative density? Does it indicate that some elements of the structure model are contradicting the experimental data? The much more plausible explanation is that both positive and negative electron density features at that level simply represent noise in the experimental data.

So what would be a correct cutoff level for interpretable electron density? Luckily, protein crystal structures provide an internal standard. If we remove a single amino acid from the 4H0H scFv model, it is expected that the recalculated omit difference maps will contain a clear indication for that residue in the form of positive omit electron density. We show the results of such a calculation in Fig. 2 (the omitted residue is the tyrosine [Y240] in the center of the image). The resulting difference electron density is equally uninterpretable to that in Fig. 1, because at the 2.5 σ level the reliable evidence in the form of positive difference density for the tyrosine is obscured by noise. If we increase the map contouring level to 6 σ, the resulting density, shown in Fig. 3, is clarified.

FIGURE 3.

Positive omit electron density (green) and negative omit electron density (red) of the Ab fragment in PDB entry 4H0H, generated by CNS. 6 σ contour level. The tyrosine in the center (Y240) was omitted from the model.

FIGURE 3.

Positive omit electron density (green) and negative omit electron density (red) of the Ab fragment in PDB entry 4H0H, generated by CNS. 6 σ contour level. The tyrosine in the center (Y240) was omitted from the model.

Close modal

It is obvious that a single amino acid not included in the model yet present in the structure is clearly detectable even at the high 6 σ difference in electron density level. If peptides were present in the crystals, it would be expected that a similar positive difference density could be observed for them as well. The area where peptide was placed in the 4H0H model is shown in Fig. 4, this time contoured at 6 σ.

Perhaps there are solvent or buffer molecules present (green spherical density features) but no density for peptide residues is visible. Features similar to those suggested by Salunke et al. to represent the peptides can be found in positive density at the 2.5 σ level in many other places throughout this structure, and it would be inaccurate to suggest that the scFv molecule binds multiple peptides. If this level of scientific proof was acceptable, many wonderful discoveries could be made from faint bands on Western blots, within margin of error changes in enzymatic activity, and in many other forms of inconclusive data.

To strengthen our argument that there is no ambiguity in the interpretation of these structural data as to the absence of peptides, we are submitting a summary (Table I) of the Protein Data Bank (PDB) validation reports for the structure models published by Dr. Salunke, and for his and our cited reference structures.

Table I.
Summary of PDB validation reports for structure models and cited reference structures
 
 

These validation results are public record (PDB validation reports, http://www.ebi.ac.uk/pdbe), and are produced by PDB. They provide the most damaging evidence by showing clearly that the peptides are structurally not reasonable, largely because they have no supporting evidence in the form of electron density.

  • 1) Based on the number of Ramachandran outliers Dr. Salunke’s peptides fall into the zeroth percentile of expected backbone geometry, whereas only a small percent of outliers is considered acceptable in a protein–peptide structure model. These peptides are in a distorted, strained high energy conformation that is simply physically improbable and would require exceptionally strong proof for their existence. We list in Table I but do not comment on the peptide models Dr. Salunke has selected as examples to defend his improbable peptides.

  • 2) The real space R-value (RSR), measuring the fit of the model to electron density, shows that there is only spurious fit and in essence 100% of the peptide residues are density outliers (RSR Z-score [RSRZ] > 2), which is again strong evidence that the implausible models were fitted into spurious density (noise).

In conclusion, these structure models, resulting from highly speculative interpretation of spurious noise, contradict the basic principles of protein structure. We do not believe that any reviewer with access to the PDB validation reports would have accepted these peptide structure models for publication. Presenting this situation as ambiguous and open to interpretation would be a scientific mistake.

Abbreviations used in this article:

CNS

Crystallography and NMR system

PDB

Protein Data Bank

RSR

real space R-value

RSRZ

RSR Z-score.

1
Emsley
P.
,
Lohkamp
B.
,
Scott
W. G.
,
Cowtan
K.
.
2010
.
Features and development of Coot
.
Acta Crystallographica Section D
66
:
486
501
.