Hello:

It is origins.com not origin.com. I made spelling mistake.  I am sorry for it.

I tried my problem. It performs better.

Thank you very much.

Regards,

Rojan

From: James Holton [mailto:[log in to unmask]]
Sent: Friday, November 19, 2010 1:05 PM
To: [log in to unmask]
Cc: [log in to unmask]
Subject: Re: [ccp4bb] origin_shift in polar space group

Ah, looks like you are running one of my scripts!  Although I called it "origins.com", not "origin.com". Someone must have renamed the file at some point.

I have a newer (aka better) version of it here:
http://bl831.als.lbl.gov/~jamesh/pickup/origins.com

which "deals" with the problem you seem to have encountered where two different chains in the "moving" model agree with the "reference" model, but with different origin shifts.  This kind of thing happens a lot more often than you might think in molecular replacement and in heavy-atom finding programs.  For example, consider trying to find two copies of the same protein in an asymmetric unit, using a crummy search model.  After the first copy has been "found", putting the second copy in the "right" place can (on occasion) have about as much dissonance as consonance with the "first" chain, and the best-scoring position of the second chain will then be found "at random" (often as not on another origin).  I thought it was interesting to know when this happens, so I left that warning message in.  However, as Ian pointed out, when you have a sliding origin, it can be hard to say exactly when an origin shift is "the same" or different.

The new version of origins.com simply picks one and only one origin shift for all chains in the "moving" PDB file and then sticks with it.  I have implemented a new approach wich uses electron density map deconvolution that seems to work pretty well, and does not rely on atom names being the same in the "reference" and "moving" PDB files.

If you do have same-name atoms in both PDB files, then the CCP4 suite program "reforigin" or the Phenix program "emma" are both much faster than my script.  However, if you get some satisfaction out of watching each origin get tried in turn, giving you a "score" for each, or if you have something like a rough chain trace and you want to compare it to a "right answer" PDB file, then origins.com might be the thing for you.

Thanks!

-James Holton
MAD Scientist

On Thu, Nov 18, 2010 at 4:26 AM, Ian Tickle <[log in to unmask]> wrote:

OK now I understand.  I couldn't find the script 'origin.com' you
mentioned in the examples directory (at least from the filename I
assume it's a script, not a MS-DOS program!), but it doesn't matter, I
see the problem now.  AFAIK there isn't a script in CCP4 that will do
what you want entirely automatically, because it's actually quite a
complicated problem in the completely general case of N molecules per
a.u., though undoubtedly it could easily be scripted for the
relatively simple case of 2 mols per a.u..

I'm assuming you don't simply want to superpose the molecules just for
structural comparison purposes, you want to superpose the entire
*crystals*, so that the calculated structure factors and hence the R
factors (values) remained unchanged for the transformed structure.
This means you can't use just arbitrary rotation/translation operators
as would be generated by superposition programs such as SSM, you have
to restrict it to crystallographically-allowed origin shifts.  There
are various programs which will do this, I wrote one called
'reforigin' but there are others which will do the same thing, and
which have been mentioned in previous postings.

So what you have to do is superpose the two 'A' molecules using
reforigin or whatever (remember, as long as it applies only
crystallographically-allowed origin shifts).  There is of course a
problem here: the chain ID 'A' is only an arbitrary label, so there's
a 50% chance that the molecule you called 'A' in structure 1 might be
called 'B' in structure 2 (and vice versa).  This means you have to
try both possibilities!  Now you see why it gets complicated in the
general case with molecules 'A', 'B', 'C', 'D' ... you have to try all
combinations!  While you are superposing A/2 on A/1 (or B/2 on A/1)
you must also transform the other chain B/2 (or A/2) using the *same*
operator (I think the program does this for you, or at least it will
print the matrix that was used for the 1st pair) - you must not
superpose it independently.

Finally you need to transform the other molecule B/2 (or A/2) in the
example above.  For this you can only use space-group symmetry
operators - you get only one chance to use the allowed origin shifts
with the first pair of molecules, after that the origin is completely
determined for the entire structure, hence only space-group symmetry
can be used to transform subsequent pairs.  For this I find it easiest
just to view the structure on the graphics, work out which is
appropriate space-group operator and apply it just to the 2nd molecule
using PDBSET.

Hope this is all clear - there are many traps here for the unwary!

Cheers

-- Ian

On Thu, Nov 18, 2010 at 10:55 AM, Rojan Shrestha <[log in to unmask]> wrote:
> Hello Ian:
>
> I am afraid that whether my problem is not clear to you.
>
> Here is brief description of the problem.
>
> When I tried to superimpose two structures having two or more copies in ASU
> for polar space group using symmetric operator, for one copies it used one
> origin and for next, another origin is used. So there is origins shift
> problem.
>
> Here is an example:
> applying  0.50  0.50 -0.69 Y,-X,3/4+Z to chain A
> applying  1.50  0.50  0.61 X,Y,Z to chain B
> WARNING: ./input.pdb chain B is on a different origin!
>
> I used origin.com to superimpose two models.
>
> Now I hope you get the insight of my problem. Do you have any idea to solve
> this problem?
>
>
> Regards,
>
> Rojan

>
>
> -----Original Message-----
> From: Ian Tickle [mailto:[log in to unmask]]
> Sent: Thursday, November 18, 2010 7:42 PM
> To: [log in to unmask]
> Cc: [log in to unmask]
> Subject: Re: [ccp4bb] origin_shift in polar space group
>
> HI Rojan,
>
> I'm not entirely clear that there is a problem.  After superposition
> any origin shift that may have been present is removed: doesn't that
> solve your problem?
>
> Cheers
>
> -- Ian
>
> On Thu, Nov 18, 2010 at 10:11 AM, Rojan Shrestha <[log in to unmask]> wrote:
>> Hello:
>>
>>
>>
>> In polar space group when the two or more copies molecules are
> superimposed,
>> the origin is shifted.
>>
>> Does anybody have the solution to tackle this problem?
>>
>>
>>
>> Regards,
>>
>>
>>
>> Rojan
>
>
>