Hi Paul,
Thank you for reading and taking time to answer my email. Please see my replies
in blue below.
Regards,
Daniel
From: Paul Emsley
Date: Thursday, 31 August 2023 at 01:46
To: Daniel Larsson ,
Subject: Re: rotamer selection
On 25/08/2023 06:56, Daniel Larsson wrote:
Hi Paul,
I have some suggestions/requests for facilitating sidechain modelling.
1. Option to have the rotamer tool default to the rotamer most closely
resembling the current rotamer (now it defaults to the most common rotamer)
That seems like a marginal improvement in the user interface.
If you want to "just tighten up the probability to the closest rotamer" then
turn on rotamer restraints so that they are used during refinement
(set-refine-rotamers 1)
The benefit of this would be that you could quickly “look-up” the closest
documented rotamer. For example, in situations when you are refining multiple
models in parallel and want to make one consistent (perhaps lower resolution)
with another. Or for analysis/documentation purposes e.g. saying “in model 1
residue X has rotamer N and in model 2 it has rotamer M”.
Thanks for the tip for how to use rotamer restraints. I have not been doing
that in COOT before, but I think I looked for it before and could not find the
feature. I will make a hot-key for that switch. Perhaps it would be nice to
have that restrain in the R/RC menu so that more people would find it?
1. Option to have the rotamer tool default to the rotamer with the highest
average map value for the atoms. Atoms closer to the Ca should be weight higher
(as an option).
This could be done quite easily. But it's getting quite close to
auto-fit-rotamer functionality.
Yes, but the distinction would be important to me. I first focus on CaBLAM and
Ramachandran outliers and after that work is done I want to disturb the
backbone as little as possible. To me, auto-fit-rotamer moves the backbone too
much and the whole mainchain geometry usually gets distorted (beyond just
tandem refine it back into shape). At first, I thought that “vanilla” auto-fit
would not move the main chain atoms and that the backrub rotamer option was
intended for this purpose, but both affect the backbone so I don’t understand
the distinction and have stopped using auto-fit-rotamer altogether.
1. Perhaps also have an option to allow for a small rotation/translation.
The devil, is as Phil would say, in the detail.
That is very true. This suggestion moot, since it would be better to just
enable rotamer restraints and refine “the normal way”.
1. Show all rotamers in the rotamer library at the same time and then you
can click on the one that you want (“bouquet” mode)
I don't see how this is an improvement over the use of the dot and comma keys.
Yes, I do use the dot and comma keys and they are great. But the long
sidechains have extensive rotamer libraries (I’m looking at you arginine) and
often many rotamers go out of the plane when you are zoomed in, so you have to
spin the structure around to see which ones can be ignored right-away and which
ones are close contenders. And when you look at them in sequence, you would
have to memorize the good ones (for example 7, 12 and 31) and then go back and
flip back and forth between them a few times to decide which one fits the
density the best while also makes biochemical sense.
Here are even more feature requests for this non-existent feature: Perhaps you
could shift/control click to deselect/select rotamers (maybe together with box
selection) and then use delete to hide bad candidates. Another suggestion which
might be easier to implement would be to color-code the options in the rotamer
list according to density fit (and maybe also clash score)? Or maybe even
color-code the rotamers in the bouquet view to take the analogy one step
further!
1. Automatically select the best rotamer in a chain/molecule considering
fit-to-density and clashes, but without touching the backbone and not doing any
refinement.
The backbone shrugs when the sidechain dances. There is no tool in Coot for
rotamer fitting that doesn't move the mainchain. It could be done quite easily,
but I am not convinced of the utility. It seems to me that it is quite possible
the the backbone atom positions have already been compromised to account for
the wrongly positioned sidechain rotamer and so without variation in backbone
orientation, the wrong rotamer is more likely to be found.
This is also true. I always do a tandem or sphere refine to a
djust the backbone after selecting rotamer. But I prefer to do it in two
independent steps to have more control. And if the density is weak, I would
probably adjust the rotamer again after that (and then point 1 above would be
helpful to quickly get an ideal rotamer) to get sound initial model restraints
for the final global minimization.
Paul.
p.s. are you iDaniel?
No, that must be another Daniel that you are thinking about
VA