Color ligands distinctly in smart initial display

[Tom Goddard]

Would be nice if ligands showed in a different color than the structure (e.g. orange) in smart initial display so they standout from the structure visually.

[Eric Pettersen]

I find that for the typically small-to-medium sized structures where people care most about ligand binding, that the ligands are pretty easily discerned. Perhaps you could provide a couple of structures where you feel the ligands are hard to distinguish and we can work from there to see if there are ways to improve those without negatively impacting other cases.

[Tom Goddard]

Any small structure with a ligand illustrates the issue, for example 5lny. I was looking at 70 of these HSP90 structures and changed the ligand colors dozens of times to more readily distinguish them from the nearby side-chains.

This request is in the same category as using separate colors for chains in small multi-chain proteins, for example 2 chain 1ah8. Currently both chains get the same color and again my first step is invariably to color the two chains differently. In past discussions I know you and Elaine seemed to think the most common scenario when looking at these small structures is comparing two or more of them so different colors are best used for the different structures. But I don't think looking at a single structure is more common.

I think using multiple colors for a single structure does not preclude using distinctive coloring for a second structure. I think multiple colors is better in the comparison case too. So if the first structure opens with protein tan and ligand yellow, the second structure would open with protein light blue and ligand orange.

My general feeling is that in the simplest and most common ChimeraX use case we make sub-optimal use of color in the initial display.

[Elaine Meng]

I haven’t found it difficult to distinguish the ligands from the protein; the presence of the ligand is the only reason the atoms are shown in that area and the ligand is (usually) not bonded to the sidechains.  Secondly, whenever I've compared multiple ligand-bound structures (often done for SFLD) it is very important to have a visual cue of which ligand goes with which structure. They aren’t just a bunch of mix and match things that I want to tell apart independently.

Chimera’s “all atom” preset shows the ligands as spheres.  That’s one way to make them more obvious, although it also makes it harder to see the chemical structure.

Another idea: Recently Greg has been interested in implementing a color lighten, darken, or tint command to make variations on existing colors, and we (TomG and I) had also discussed something similar in the past.  Maybe ligand carbons could be a lighter version of its model or chain color.  However, if your goal is to have the ligand “stand out” starkly like you might want for presentation images, that won’t help.

I agree it is useful to see that there are multiple chains when you open a single structure.  However, if we used the “bychain” coloring on each successive structure I think it would be bad that all the chains A of different models would be the same color, all the chains B the same color, etc.  Maybe that isn’t what you meant to suggest, though.  I was thinking the “bypolymer” coloring might be good because it would be different for different proteins, but then all the homomultimers and different complexes of (exactly) the same molecular entity would not be different colors.

[Eric Pettersen]

Like Elaine, I have also found it important to have the ligands colored similarly to their structure when examining similarities and differences of binding pockets when structures have been matched on each other.

I feel we have had this discussion many a time. We make the single-structure case slightly less optimal so that the multiple-structure case isn't made significantly less optimal. If we somehow psychically knew that the user was not going to compare a newly opened structure to others in that session, then what you suggest would clearly be the way to go (well, mostly :-) ). In ChimeraX, it's even finer grained then in Chimera 1 in that for structures with 5 or more chains we depict that with individually colored chains based on the guess that such structures are less likely to be compared to other structures than their smaller counterparts.

It looks like we are never going to agree on the "right" initial structure depiction, so it would seem that there need to be preferences to offer control over the depiction: whether to individually color chains; whether that depends on the number of chains (and how many); should they be colored by sequence; should ligands have "stand out" colors or be colored like their structure (byhet) or by element, etc.

If we can agree to resolve it in this fashion, I'll put it on my to-do list.

--Eric