Progress report July 5 - September 7, 2023

Tom Goddard
September 7, 2023
Group meeting

Recent work

Demonstration of LookSee Quest VR
IRAK4 dimer with inhibitor (PDB 8ATL).

LookSee example use

Setup steps for LookSee example

Discussion of ChimeraX Reviewer Comments
  1. In Table 1, The reference to the Qscore is not included; it should refer to current reference 34, but will likely have to be renumbered.

    Answer:
    Scooter has added the QScore reference.

  2. In the caption to Table 1, I suggest it to read "ChimeraX extensions described in this paper. The developer who ported the extension to ChimeraX is listed; the original developers of the external software are referenced in Column 1.", if this is the case.

    Answer:
    We improved the table 1 caption to now read: "ChimeraX extensions described in this paper. The developer who implemented the extension is listed and the first column cites the publication of the extension or original algorithm."

  3. In the AlphaFold section (p. 3-7) there is a very relevant assessment of the results of AlphaFold modeling with respect to cryoEM structures. However, it is difficult to relate some of the figures to each other. These include the relationship of Fig 1a to 1c, 1e and 1g; Fig 2a, 2b and 2c to 2f; and fig 4a and 4e to 4b-d. The portions corresponding to parts of larger structures might be indicated by boxes or some other means.

    Answer:
    We will carlify the wording of figure captions making them more self-contained so they convey the purpose of the figure without having to read the main text of the paper. In figure 1 we can improve the c) description to say it does not match the cryoEM data, and in g) we can say it is the same as e but shows just the domains observed in cryoEM. In figure 2 we can explain f) better saying it shows the 5 ALphaFold predictions are very similar to one another, and refer to it as a "single monomer" of the dimer prediction. In figure 4 we can improve the bold title to say something like "Alphafold predicted insulin binding is at a secondary binding site". Also might say a-c are human to contrast with d mouse.

  4. In preparing figures for publications, it would be convenient if plots such as Fig 1d and 1f or 2d and 2e were able to be directly exported with readable scales and fonts for inclusion in manuscripts.

    Answer:
    We interpret this as a request that ChimeraX software be able to export the plots. It can currently export the PAE plots as a high-resolution PNG image and the ChimeraX colorkey command can make a suitable color key image that the user can compose into a figure. We agree being able to export a ready-made publication figure would be a nice feature.

  5. In Fig 3 caption, it is said that the final cryoEM structure is not shown, but fig 3b and 3d appear to show the final cryoEM structure.

    Answer:
    The "not shown" text is describing panel a), not panels b) and d). The b) and d) descriptions say they are the final cryoEM structure. It seems the reviewer missed that the "not shown" was in the a) description.

  6. In the Model Assessment discussion (p. 10-11), Qscore evaluates main chain, ligand, water, and nucleic acid densities as well as side chains. It is also part of the Validation Report in PDB deposition. The normalization of Qscore to local resolution may indeed help distinguish modeling errors from map resolution issues, but I suspect they are related; sometimes models may just fit more poorly in maps of lower resolution.

    QScore section text:
    "QScore approximates the side chain density by placing Gaussians at atom positions and correlates this predicted density to what is observed in the map. It only considers density closer to the assessed atoms than to any other atom. The Gaussian width is chosen to match very high-resolution maps (< 1.5A), so the value that is considered a good fit depends on the map resolution. This measure is the most widely used per-residue map fit score and is reported by the EM Databank for fit models. The ChimeraX extension can compute it for 1000 residues in 10 seconds and plot the scores. Clicking on a residue in the plot shows the corresponding residue and nearby map contour surface in the 3D graphics window."

    Answer:
    We will add text saying that QScore works on atoms other than amino acids. I checked that Tristan's QScore does work on nucleic acids, ions and ligands.

  7. In Fig 7 caption, the drug is inconsistently referred to with respect to the main text above it.

    From the text:
    "...aid understanding in vivo mechanisms such as the anticancer drug homoharringtonine."

    Caption:
    "Figure 7: Spatial arrangement of ribosomes seen in untreated cells and cells treated with anti-cancer drug homoharrington, with colors indicating the ribosome conformational state (Figure 4 of Xing et al. 202338). Ribosomes are more dispersed in drug-treated cells. "

    Answer:
    We will correct the caption to say homoharingtonine.

  8. It would be nice if UCSF ChimeraX could prompt the users using the individual packages in the toolshed to cite the primary papers in addition to the UCSF ChimeraX papers.

    Answer:
    This is already supported. ChimeraX extensions on the Toolshed often do log a message asking to cite their publication when used. Other extensions include in the user interface panel the citation. It is up to the extension developer to add this if they wish.

  9. CryoEM is abbreviated in the introduction as "cryo-EM" but used as "cryoEM" in the manuscript. CryoET should be abbreviated similarly.

    Answer:
    We will use consistent hyphenation. Publications use both variants, although the big name cryo-EM labs seem to prefer the hyphen so we will use a hyphen.

  10. Figure 1 and 2 have some overlap. I would recommend extending Figure 2 with the latest advances in multimer prediction.

    Answer:
    ChimeraX used the current AlphaFold version 2.3 for predicting the insulin dimer. While there are other prediction methods we believe we have used the latest state-of-the-art method, and of course the paper is limited to what is available in ChimeraX.

  11. ColabFold works only for users with a google account.

    From the text:
    "A new AlphaFold prediction (monomer or multimer) can be run on Google Colab servers by pasting the desired sequence(s) into the ChimeraX AlphaFold tool."

    and two paragraphs further down

    From the text:
    "A maximum total sequence length of about 2000 residues can be predicted using a free Google Colab account. "

    Answer:
    We can add to the text when Google Colab is first mentioned that it requires a free account.