Re: [ccp4bb] AW: Going back to Coot 0.8
Dear Eike,
This is not about me trying to convince you to change. I was pleased
with Coot 0.8.9.2 when I released it and I am pleased that you like it
too. I was trying to express that *I* think that the refinement in
0.9.x is across the board better and I am not seeing the problems that
you see.
Instead, this is about me trying to find out what you (and others)
think is going wrong. This is about turning your email into a useful
bug report. What I've got now is "it does not work any more and I
don't like it" - and that is not helpful.
You mentioned that you had problem with a ligand complex. I tried to
investigate further by asking you if you were using Acedrg-Coot. You
didn't answer.
You suggest that atom clashes may be a problem. I tried to investigate
further by asking you if you noticed a problem with Coot's description
of non-bonded contacts. You didn't answer.
To be clear, these were not rhetorical questions, they were an attempt
to address the problems - or at least begin to do so.
What I need is an example of *what* is going wrong. Just one example
would be a start. It seems to me that if the problem is as pervasive
as you seem to think it is, then it shouldn't be difficult for you to
conjure up a problematic case. This is the sort of thing that I want:
This model is based on 1xyz from the PDB, except that I've adjusted
the model to create a problem. See the region of A501 - you can
see that that model and the map do not correspond. It is hard for
me to fix it in 0.9 because when I do X, Coot does Y, but in 0.8.9
all I had to do was A, B and C.
It might be the case that Coot can trivially solve the problem, but
you don't know about it - and that might be because I have
insufficiently explained how to do it. Or made the means by which you
can do so too opaque. Or maybe you have to type something. At the
moment I have no clue.
Normally, I would have taken to private correspondence by now. I
recognise that your primarily interest is the latest and greatest of
the 0.8 series, and you may think that this may not be worth your
effort, so if you don't do this perhaps some of the other frustrated
Coot users may do so? Hence the public continuation of our discussion.
Regards,
Paul.
On 07/09/2020 10:28, Schulz, Eike-Christian wrote:
Dear Paul,
Thanks for the detailed response. The reason for me being a bit annoyed is really
simple: basic model building did work well in coot 0.8.x and (for me and others)
it does not work anymore in coot 0.9. Because I like coot (and because most
computer problems are usually 30cm in front of the screen) I have tried to get
used to it for a few month, but on the weekend I have switched back to coot 0.8 *
–> problem(s) solved, what a relief.
I realize that I have not been particularly helpful in bug-fixing, but the
changes appear to be rather comprehensive and thus I can only comment on the
user experience. This is (naturally) a subjective experience but what I hear
from my colleagues tells me that I am not alone with my frustration.
Try to look at it from my point of view:
- the previous version of coot does what I want,
- I see few advantages of the new version,
- the basic functionality (that I am used to) appears to be gone in the
new version.
Since I am not convinced - why should I switch ?
Best regards,
Eike
* thanks to Bill Scott for the pre-compiled mac-binaries
http://scottlab.ucsc.edu/xtal/wiki/index.php/Stand-Alone_Coot
-Original Message-
From: CCP4 bulletin board on behalf of Paul Emsley
Reply to: Paul Emsley
Date: Saturday, 5. September 2020 at 02:05
To: "CCP4BB@JISCMAIL.AC.UK"
Subject: Re: [ccp4bb] AW: [EXTERNAL] Re: [ccp4bb] AW: Going back to Coot 0.8
Dear Eike,
Yes, it helps a little.
Changing weights on switching refinement maps is inconvenient, but not a
point of failure.
Alternative conformations may need some assistance - pin an atom using
anchoring.
When Coot starts a refinement it tells you in the terminal about the bad
bonds and non-bonded contacts that it found in the model. Do you
perceive that to be wrong or missing interactions?
There are many ways to skin a covalent ligand complex. I use the
Acedrg-Coot interface (as described in my blog). Do you?
If you wish, you can calculate an average map, the tool for that can be
found in Calculate -> Map Tools -> Average Maps...
Perhaps it would be useful to have an alert "You're not looking at the
map into which you are trying to fit the model" - maybe the background
should go pink - or something.
I feel that you are annoyed but I still don't understand why. As Phil
used to like to say, the devil's in the details.
I see that previously I didn't answer your question. So let me correct
that now:
(i) You can turn on interactive contact dots ("Contact Dots On") using
Refinement Tools from Curlew. This i
Re: [ccp4bb] AW: Going back to Coot 0.8
Dear Paul,
Thanks for the detailed response. The reason for me being a bit annoyed is
really simple: basic model building did work well in coot 0.8.x and (for me and
others) it does not work anymore in coot 0.9. Because I like coot (and because
most computer problems are usually 30cm in front of the screen) I have tried to
get used to it for a few month, but on the weekend I have switched back to coot
0.8 * –> problem(s) solved, what a relief.
I realize that I have not been particularly helpful in bug-fixing, but the
changes appear to be rather comprehensive and thus I can only comment on the
user experience. This is (naturally) a subjective experience but what I hear
from my colleagues tells me that I am not alone with my frustration.
Try to look at it from my point of view:
- the previous version of coot does what I want,
- I see few advantages of the new version,
- the basic functionality (that I am used to) appears to be gone in the
new version.
Since I am not convinced - why should I switch ?
Best regards,
Eike
* thanks to Bill Scott for the pre-compiled mac-binaries
http://scottlab.ucsc.edu/xtal/wiki/index.php/Stand-Alone_Coot
-Original Message-
From: CCP4 bulletin board on behalf of Paul Emsley
Reply to: Paul Emsley
Date: Saturday, 5. September 2020 at 02:05
To: "CCP4BB@JISCMAIL.AC.UK"
Subject: Re: [ccp4bb] AW: [EXTERNAL] Re: [ccp4bb] AW: Going back to Coot 0.8
Dear Eike,
Yes, it helps a little.
Changing weights on switching refinement maps is inconvenient, but not a
point of failure.
Alternative conformations may need some assistance - pin an atom using
anchoring.
When Coot starts a refinement it tells you in the terminal about the bad
bonds and non-bonded contacts that it found in the model. Do you
perceive that to be wrong or missing interactions?
There are many ways to skin a covalent ligand complex. I use the
Acedrg-Coot interface (as described in my blog). Do you?
If you wish, you can calculate an average map, the tool for that can be
found in Calculate -> Map Tools -> Average Maps...
Perhaps it would be useful to have an alert "You're not looking at the
map into which you are trying to fit the model" - maybe the background
should go pink - or something.
I feel that you are annoyed but I still don't understand why. As Phil
used to like to say, the devil's in the details.
I see that previously I didn't answer your question. So let me correct
that now:
(i) You can turn on interactive contact dots ("Contact Dots On") using
Refinement Tools from Curlew. This is a validation tool however and
doesn't describe what is going on with the refinement energy inside
Coot's head. I have it turned on by default for myself, FWIW.
(ii) There is no way to turn on "Old Style" refinement in 0.9 - although
"proportional editing" in 0.9.1 will bring some of that non-refined
initial displacement of surrounding atoms [1]
(iii) 0.8.2 is the best release for the 0.8.x series. I did quite a bit
of work on 0.8.3 but I never released it. Most (not all) of the updates
have been folded into 0.9.x.
Paul.
[1] it was seeing that Tristan had added to ISOLDE something like the
tool I had intended that that actually made me write it for Coot [2] -
it is a useful technique when shoving around domains in cryo-EM.
[2] That and watching Blender tutorials.
On 04/09/2020 19:04, Schulz, Eike-Christian wrote:
> Dear Paul,
>
> thanks for your comprehensive response in spite of my rather general
comment. And thanks for listening! I hope you see this as constructive
criticism and not only as blunt complaining (a national tradition ;) ):
>
> Instead of a screencast I'd like to give at least a few more details on
the trouble on my end.
>
> The problems mainly arise when:
>
> - switching maps e.g. from 2Fo-Fc to Fo-Fc, or omit maps (X-ray
restraints ?)
> - modelling alternative / low occupancy conformations
> - modelling covalent ligand complexes (atom clashes ?)
> - parts of the solution are in either of the maps (maybe map averaging
would help?)
>
> --> which is practically all I am doing, and the result is often a model
next to but not inside of the density. I am a long-term practitioner of
ctrl-drag, but with little help in these cases.
>
> With respect to sophisticated solution and the "method developers’
intention", I would like to answer with some experience from our own serial
crystallography method development.
>
> For even distribution of micro-crystal slurries, we developed a
custom-multichannel pipette adapter, went through different iterations, adapted
and improved the design added more features optimized parameters and it now
works nicely as intended. However, users rather like to use a single empty
pipette-ti
