[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ccp4bb]: Summary: Amore Rotation Function Scoring?
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dirk Kostrewa wrote:
>
> *** For details on how to be removed from this list visit the ***
> *** CCP4 home page http://www.ccp4.ac.uk ***
>
> Dear CCP4 users & staff,
>
> on September, 23rd, I've posted the following message on the CCP4BB:
>
> "The Amore cross-rotation function basically calculates a correlation
> coefficient between the observed and calculated Patterson function (CC_P).
> However, the output of the cross-rotation search is for some dubious reason
> sorted on the correlation coefficient between calculated and observed F
> (CC_F). This doesn't make much sense to me for the following reasons:
> (1) The search function is the CC_P, thus, from a methodological point of
> view, the output should be sorted on this value and not on something else.
> (2) Both, the calculated F and I of the model only make sense after it has
> been correctly positioned, which is not the case in the cross-rotation
> search.
> (3) Accordingly, the signal-to-noise must be much better for CC_P than for
> either CC_F or CC_I. To illustrate this, I have run a cross-rotation search
> with the refined protein-only model of the A. niger phytase (Kostrewa et al.,
> NSB, 4, 185ff, 1995) against its observed data. The top 10 of the amore
> cross-rotation output looks like this (I've removed the TX,TY,TZ columns for
> better readability):
>
> ITAB ALPHA BETA GAMMA CC_F RF_F CC_I CC_P Icp
> SOLUTIONRC 1 3.28 85.77 237.92 27.9 55.3 42.8 26.8 1
> SOLUTIONRC 1 117.85 90.00 58.64 22.2 57.2 34.3 16.3 2
> SOLUTIONRC 1 90.57 80.41 235.17 18.2 58.3 26.3 5.6 3
> SOLUTIONRC 1 60.20 85.07 240.47 17.9 58.5 25.9 4.9 4
> SOLUTIONRC 1 22.57 57.12 223.13 17.9 58.5 26.6 4.1 5
> SOLUTIONRC 1 47.85 86.10 237.71 17.8 58.5 25.8 5.2 6
> SOLUTIONRC 1 87.65 60.22 71.67 17.8 58.4 25.9 4.4 7
> SOLUTIONRC 1 80.37 85.82 235.99 17.7 58.4 25.1 4.5 8
> SOLUTIONRC 1 44.86 24.72 48.00 17.7 58.5 26.0 5.6 9
> SOLUTIONRC 1 41.18 58.25 87.29 17.7 58.4 25.7 6.4 10
>
> Interestingly, the correct top peak appears to be also the top peak in CC_F
> and CC_I. However, as you can clearly see, the signal-to-noise ratio is MUCH
> better for CC_P. Now, imagine that you do not have a perfect search model. In
> this case, I think, the chances to find the correct peak are much poorer if
> the output is sorted on CC_F rather than on CC_P. I don't know what you other
> users of CCP4 think about this, but I would strongly prefer a sorting on the
> real search function values rather than on something else in order to get the
> best chances to find the correct molecular replacement solution.
> Unfortunately, CCP4 Amore apparently does not give the user the choice on
> which values he/she wants to sort the output. Thus, the request from my side
> to the CCP4 developers is to give the user the choice on which values the
> output should be sorted, and to set the sorting on CC_P as the default, and
> not the sorting on CC_F."
>
Sorry not to have answered - I really cant quote chapter and verse to
prove or disprove the most reliable indicator..
However first point: the CC_F is actually the CC( F-<F> ) - equivalent
to cc on E**2 -1 for normalised amplitudes.
Jorge Navazadid tests, and found that all indicators detect strong
signals, but that this one was more likely to detect weak signals for
low homology models than the CC_P. It also seemed less vulnerable to
missing and unreliable data.
In fact the most statistically valid test is that used by BEAST but it
is much slower, and since we only want a set of 50 or so trial
orientations from the rotation function for further analysis in the
translation, the CC_F seemed the most sensitive. It really doesnt matter
whether the correct orientation is 11st or 17th - providing it is
present..
Eleanor Dodson