[BiO BB] a validation study

Dan Bolser dmb at mrc-dunn.cam.ac.uk
Mon Dec 6 21:20:57 EST 2004

On Mon, 6 Dec 2004, l x yi wrote:

>I have developed a new method for searching protein
>banks with PSSM (position-specific matrices). To do a
>validation analysis of the method on real protein
>sequences, I need to decide 

Their are papers which ask the same questions (I don't have references to
hand), so you could try to extend the analysis in them to cover your

As a gold standard lots of people pick SCOP as a definition of true (but
very distant) relationships. You can check your results at family /
superfamily and fold levels (folds are considered to be non-homologous but
structurally similar protein domains).

Generally there are two criteria of alignment, quality and length. If you
are trying to do homology modeling you prefer length, pure fold
recognition you could go for quality. 

I am sure lots of people can add much more details to this very shakey

All the best

>-- a selection of some position-specific matrices for
>major protein domains
>-- a databank of single sequences
>-- a gold standard of which sequences truly belonged
>to which protein domain
>Could anyone suggest some reliable sources of
>references as to how to conduct this validation
>analysis? Any suggestions are appreciated. 
>Do you Yahoo!? 
>Take Yahoo! Mail with you! Get it on your mobile phone. 
>BiO_Bulletin_Board maillist  -  BiO_Bulletin_Board at bioinformatics.org

More information about the BBB mailing list