I have three published alignments, and I want to use my models generated
by Modeller to qualitatively estimate the goodness of the different alignments.
I have made 100 models of my unknown sequence to the target structure using
each of the three alignments.
I then binned the objective energy output from the standard Modeller run
(see panel A in attached pdf). I also binned the final C-alpha to C-alpha
RMS for the generated models to the known homologous structure (see
panel B in attached pdf).
Is it true that the energy function is not the best objective measure because different
alignments will have different numbers of constraints (i.e. one alignment may
correspond to 100 amino acid pairs, whereas a shifted alignment may only have
98 amino acid pairs between the known and unknown sequences.) Can you
easily normalize to account for this?
What is the best way to do this? I would like to say that Alignment 1 below is best
but I am hesitant.
Hopefully this question is understandable.
Thank you,
Michael
--------------------------------------------------------------------------------------------------
Michael Grabe, Ph.D.
Post-doctoral Fellow
Howard Hughes Medical Institute
University of California, San Francisco
533 Parnassus Ave.
San Francisco, CA 94143