I noticed this kind of thing myself a long time ago, and wondered what
refmac was doing to make things "worse", so I let it keep going. And
going and going. I was delighted to "discover" that although R and/or
Rfree could rise over up to hundreds of cycles, it almost invariably
turns around again,
Hi AR
Please define what you mean by 'over-refinement' as it's not a term I use:
does it mean 'convergence', or 'over-fitting', or 'over-optimisation'
(whatever that means) or something else?
If by "LLG is stabilized" you mean it has converged then I agree that's a
possible stopping criterion, bu
refinement 20 cycles is usually enough
(but more cycles shouldn't hurt).
Cheers,
Robbie
Date: Fri, 26 Aug 2011 20:29:59 +0530
From: proteinchemistr...@gmail.com
Subject: Re: [ccp4bb] number of cycles in refmac
To: CCP4BB@JISCMAIL.AC.UK
Dear Dr Ian
from your argument i could not under
Dear Dr Ian
from your argument i could not understand how many cycles to refine before
submitting the coordinates to the PDB. what is the upper limit 100 or
thousand or million according to my understanding, its more logical to
stop the refinement when over refinement is taking place (when R
Frank,
Point #1 - fair point; the reason Rfree is popular, though, is because it
> is a *relative* metric, i.e. by now we have a sense of what "good" is. So
> I predict an uphill fight for LLfree.
>
Why? I don't see any difference. As you say Rfree is a relative metric so
your sense of what 'g
> Point #2 would hold if we routinely let our refinements run to
> convergence; seems common though to run "10 cycles" or "50 cycles" instead
> and draw conclusions from the behaviour of the metrics. Are the conclusions
> really much different from the comparison-at-convergence you advocate?
> Wh
Hi Ian
(Yes, your technical point about semantics is correct, I meant
over-fitting.)
To pin down your points, though, you're saying:
1) Don't use Rfree, instead look at LLfree or Hamilton Rfree.
2) Compare only the final values at convergence when choosing
between different parametriz
Hi Frank
> This is self-evident; what is not obvious is why the target function
should be having the final word. Wasn't the word "over-refinement"
introduced to describe exactly this: that the target function was wrong?
I assumed people were confusing 'over-refinement' with 'over-fitting'; the
Hmm, I used to think I understood this, but I'm feeling a bit dim right
now.
On 25/08/2011 11:07, Ian Tickle wrote:
Since the target function in MX refinement is the total likelihood
(working set + restraints), there's no reason whatsoever why any
another function, such as Rfree & LLfree, shou
TIm,
I dare say that the goal is to get phases which match as good as
> possible with what is inside the crystal. If this coincides with
> maximising the likelihood, why don't we run refinement until the LL
> stabilises?
>
That's exactly what you should do: any optimisation procedure attains the
Hello Tim
It was in one or two versions and I did not get consistent results. However
code is there and I can activate it if you want. If you know what criteria you
would like to use I can code that also.
In some cases it happens that R/Rfree go up and then they start coming down. It
may be ca
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Hello Ian,
I dare say that the goal is to get phases which match as good as
possible with what is inside the crystal. If this coincides with
maximising the likelihood, why don't we run refinement until the LL
stabilises?
@Garib: I have seen runs wher
Dear Tim
At the moment there is no option to stop refmac prematurely. I can add if it is
necessary. I can only give my experience.
After molecular replacement before running ARP/wARP or buccaneer I usually run
60-100 cycles of refinement with jelly body with sigma set to 0.01.
Then automatic mo
Hi Tim
The answer is a definite NO, the goal of refinement is to maximise the
likelihood from the working set and the restraints. It is definitely not to
optimise Rfree or LLfree. The correct value of the latter is whatever you
get at convergence of refinement, i.e. at maximum of the likelihood,
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Dear all,
especially at the beginning of model building and/or at low resolution
both Rfree and "-LL free" as reported in the refmac logfile show a
minimum at a some cycle before rising again.
I am certainly not the only one tempted to first run refm
15 matches
Mail list logo
