Dear Thomas, Thank you for pointing this out. I'm pleased to say that there is no significant change to the scatter plots of MLAD versus correlation coeeficients when using your change. So you should still be fine using the current version of find_alt_origin_sym_mate. The next release of PHENIX will use the changed version of the MLAD formula. This will eventually be documented online. Many thanks, Robert On Feb 21 2012, Thomas Holder wrote:
I also have a question regarding MLAD score.
Why do you take the sum of absolute dr_i components? This is not orientation invariant. The attached screenshot is like figure 3 from the article, but rotates the system from 0 to 90 degrees.
I think this:
|dr_i,x| + |dr_i,y| + |dr_i,z|
needs to be replaced by this:
|dr_i|
right?
Regards, Thomas
On 02/21/2012 01:31 AM, R.D. Oeffner wrote:
Dear Bryan,
The factor 0.9 is a purely empirical term. It serves two functions. One is to soften the minimas when atoms coalesce and to also prevent taking the logarithm of a negative number. The closer it gets to zero the more accurately it pinpoints the global minimum when the MLAD score is used for identifying the best superposition betweent two molecules. But if it becomes two small chances are that the landscape of MLAD values will be dotted with many very deep local minima which might lead to wrong results.
The square boxes under the min, max symbols are typos to be ignored, sorry.
There are currently no plans to publish further details about find_alt_origin_sym_mate.
Regards,
Robert Oeffner
-- Robert Oeffner, Ph.D. Research Associate Department of Haematology, University of Cambridge Cambridge Institute of Medical Research Wellcome Trust / MRC Building, Hills Road, Cambridge, CB2 0XY www-structmed.cimr.cam.ac.uk, tel:01223763234, mobile:07712 887162