Escrever um comentário

Summary

The major goal of this paper is to use “guided Alphafold 3”, an exciting new way to do ensemble refinement, to quantify how widespread alternative conformations are in β2-microglobulin. The major strength of the paper is the detailed method application to a specific system. If the major goal is, however, to determine whether the alternative conformations exist, simple occupancy refinement with the identical set of alternative conformations in each dataset could likely arrive at an answer much more simply (and would be a good baseline regardless). With comparison to the baselines, it will also be clearer whether this new procedure is re-discovering the same conformations or sampling new ones. Regardless, this type of exploration is an exciting direction for the field and this method has large potential to discover unmodeled alternative conformations.

Major points

1. The "revisiting" of an old set of PDB depositions is interesting, and we think is more warranted as we've developed more powerful tools for analyzing the data.

   1. However, it seems like for this case, looking at the e- density is a better proxy for this particular task. We wish that the figures displayed the density more clearly, as looking directly at the density in real space may provide clear support for the existence of altlocs

2. The claim that C 1 2 1 (C2) and I 1 2 1 (I2) are different in their symmetry is false. These two space groups are only different in convention, and are both representations of space group 5. The conclusions of the results section that refer to this need to reflect this, and the authors might benefit looking into the changes in the unit cell, potentially due to differences in packing. The examples that are truly polymorphs deserve more focus, in this case.

3. map_box seems like an important caveat for claim at the very end, of it being a general methodology. It seems that the approach detailed here requires prior knowledge of existing heterogeneity at a location. It would be useful if the authors speculated on ways that this process could be automated to truly uncover useful information by re-refining old data.

4. Fig 2E, RSCC seems nearly the same between structures in the two space groups? Caption explanation is not supported by the actual figure

5. How do other methods that aim to recover multiple conformations from the data, such as ensemble refinement, perform here?

Minor points

1. How is similarity for color measured? It would be useful to explain this in the text (It might be there, I haven't found it yet)

2. Fig 4 caption talks about refined ensembles, which we presume is the guided ensemble, but this term is not used elsewhere in the paper

3. Fig 2 panel lettering is not correctly used in the caption

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.