Dear Jon,
I don't think we have any disagreement. I just wanted to emphasize
that you should have your plan thought out at the start. You may have
decided that you will compare your ligands shape with your map using a
Real Space R Factor. If you don't like that number the fault isn't in
the RSR but in your density.
Of course you could have decided ahead of time to use a correlation
coefficient. Or you could have planed to calculate both and defined
some weighting scheme to use the two in making the decision. The error
comes in if you decision tree depends on how well the analysis justifies
your desired outcome.
Like your test set, how you make your decisions should be kept
isolated from your actual analysis. The power of the human mind to bend
its choices toward a rewarding outcome, even unconsciously, is enormous.
I was trying to stay away from the particulars of what that decision
tree would look like. That topic has been discussed many times on the
BB. I certainly agree with my good friend Blaine Moors that the Fo-Fo
map is the gold standard for deciding if something bound and gives a map
that is unbiased by modeling. In addition, Fo-Fo maps between the
crystals of varying occupancy, even very small changes in occupancy, are
surprisingly informative. They tend to be highly isomorphorus and
provide direct information for deconvoluting multiple conformations
which is vital in partial occupancy binding.
Dale Tronrud
P.S. Changing the contour level does not change the map. That is simply
a representation issue due to the difficulty of presenting all the
information in a map. Sharpening or blurring a map makes a new map, and
since the sharpening factor is a continuous number that dial wheel
creates an infinite number of different maps. If your only means of
selecting which one is "best" is how well the map fits your ligand, that
map can't be used to justify your interpretation.
This could be made rigorous by, for example, deciding on the factor by
looking at the quality of the map in some uncontroversial region -- If
you decide on the means of choosing that region ahead of time.
On 11/24/2020 8:20 PM, Jon Cooper wrote:
Hello Dale, the statistical rigour you describe is, of course,
excellent, but in a learning environment, if someone gets a negative
result, you have to go into overdrive to check that everything has been
done correctly, since there is a fair chance that human error is the
cause. It may be a terrible practice, but it would seem to be an
important part of the process? Even as a relative newcomer to the field
(well, since the mid-80's ;-) I have seen many people getting nothing in
their initial difference maps, even if the ligand is there. Frequently
it was just the contour level being too high and, depending on how far
back you go, the solution varied from showing someone how to roll the
mouse wheel in Coot to having the map recontoured at a computer centre
200 miles away and posted back on a magnetic tape, which took about 10
days - a timescale on which some people just gave up and did something
else! I can't help thinking it would be a shame to robotically accept
every negative result at face value, not least if you're doing something
important like curing a pandemic. However, back to the original question
which I think was whether polder map coefficients could be used as
refinement targets and I think the answer to that one is probably 'no',
at least in the X-ray field ;-)
Best wishes, Jon Cooper
-------- Original Message --------
On 24 Nov 2020, 16:02, Dale Tronrud < de...@daletronrud.com> wrote:
Hi,
To me, this sounds like a very dangerous way to use this tool decide
if a ligand has bound. I would be very reluctant to modify my map with
a range of arbitrary parameters until it looked like what I wanted to
see. The sharpening and blurring of this tool is not guided or limited
by theory or data.
As you describe it, your choice of map is driven by its agreement
with your ligand, and the proper way to make this decision is the other
way around.
The original poster has the problem that their density does not have
the appearance they desire. They have chosen to run around trying to
find some way to modify the map to get a variant that does. This is a
terrible practice, since the final choice of map is being made in a
fashion that is dominated by bias.
I have no idea what sort of "structural characteristics" have
convinced this poster of the presence of their ligand despite the
absence of clear electron density. What other evidence does a
diffraction pattern give? The map is your best and only source of
information about your structure that you can get from the diffraction
pattern. (Mass spec and other experimental techniques could, of course,
be applied.)
I think we, as a community, could learn a few things from the
vaccine trial studies that are so much in the news now. In a modern
clinical trial, to avoid bias in the interpretation of the results, all
of the statistical procedures are decided upon BEFORE the study is even
began. This protocol is written down and peer reviewed at the start.
Then the study is performed and the protocol is followed exactly. If
the results don't pass the test, the treatment is not supported. There
is no hunting around, after the fact, for a "better" statistical measure
until one is found that "works".
This way of handling data analysis in clinical trials was adopted
after the hard lesson was learned that many trails could be reproduced,
their results were not.
I would recommend that you decide what sort of map you think is the
best at showing features of your active site, based on the resolution of
your data set and other qualities of your project, before you calculate
your first Fourier transform. If you think a Polder map is the bee's
knees then calculate a Polder map and live with it. If you are
convinced of the value of a FEM, or a Buster map, or a SA omit map, or
whatever, calculate that map instead and live with it.
If you have to calculate twenty different kinds of maps, with
varying parameters in each, before you find the one that shows the
density for your ligand; it probably didn't bind.
Dale Tronrud
On 11/24/2020 5:35 AM, John R Helliwell wrote:
> Dear Nika,
> A tool I am gaining experience with, but for a challenge like you
> describe, may help:-
> In Coot>Calculate you see “Blurring/Sharpening tool”. You are
> presented with a choice of electron density map (here you would
select
> your Fo-Fc). There is then a slider tool, to the left and to the
right,
> and you can see the impact of negative or positive B factor on
your map.
> Blurring, slide right, may assist your density continuity versus
> Sharpening, slide left, which may assist the detail of your map. The
> logic of the tool is that your diffraction data, and of the Fo-Fc
> differences, can be fine tuned, in or out.
> Best wishes,
> John
>
> Emeritus Professor John R Helliwell DSc
>
>
>
>
>> On 24 Nov 2020, at 11:29, Nika Žibrat <nika.zib...@ki.si> wrote:
>>
>>
>>
>> Hello,
>>
>>
>> I have a question about protein-ligand, of which ligand displays an
>> ambiguous electron density. I am solving a structure of protein with
>> ligand which was obtained via soaking. Structural characteristics
>> indicate the ligand is present however the electron density is quite
>> vague and too small for the size of the whole ligand. I did a Polder
>> map which showed much larger area of green density. After
insertion of
>> my ligand into the green density in Polder I ran phenix.refine and
>> there is a lot of red on the spot where the ligand is which was
to be
>> expected. This leaves me wondering how, if even do I incorporate the
>> polder map data into my refine input.
>>
>>
>> My question is, how do I continue refining and validating the
>> structure in this case?
>>
>>
>> Thank you,
>>
>>
>> Nika Žibrat
>>
>>
>>
>>
------------------------------------------------------------------------
>>
>> To unsubscribe from the CCP4BB list, click the following link:
>> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>
>>
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>>
>>
>
>
------------------------------------------------------------------------
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>
> <https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>>
>
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>
This message was issued to members of www.jiscmail.ac.uk/CCP4BB
<http://www.jiscmail.ac.uk/CCP4BB>, a mailing list hosted by
www.jiscmail.ac.uk <http://www.jiscmail.ac.uk>, terms & conditions
are available at https://www.jiscmail.ac.uk/policyandsecurity/
<https://www.jiscmail.ac.uk/policyandsecurity/>
------------------------------------------------------------------------
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
<https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
