In article <3tug4i$7jg at nexus.uiowa.edu> kpmurphy at blue.weeg.uiowa.edu (murphy kenneth p) writes:
>> In article <3tucm1$gg6 at agate.berkeley.edu>,
> Louis Hom <lhom at OCF.Berkeley.EDU> wrote:
> >I was just wondering if that paper by Rose et al. regarding The Problem had
> >come out yet. I seem to recall someone predicting its publication sometime
> >last month or this month. Medline showed nothing.
> >--
> >______________________________________________________________________________
> >Lou Hom >K'93 I hear the American Enterprise Institute
> >lhom at ocf.berkeley.edu says it's true.
> >http://www.ocf.berkeley.edu/~lhom/>> Yes, it's in the latest issue of Proteins as the lead article.
>
so what did you gues think. Is in time to try to join the AVP
beachvolleyboll teams, or should I continue my science carrier ?
Well I'm starting my ow lab next month so I guess that is my answer.
However if you calculate the accuracy of secondarystructure prediction
in the 5 proteins reported is 92 % (and if you do not include apo-myoglobin
it is it is 94 %). Which is more or less as good as you get if you compare
for instance between procheck and dssp for the same protein, and better than
between highly related structures. But this is for only four (different)
proteins.
On the other side the packing of secondary structure elements is not that
impressive. Looks pretty good for the fragments in fig 3 (1fb) and for
2 out of three 50 aa fragments of b562. but pcy, mbo and eglin does not
look that good (when you take into aspect that you have nearly perfect
secondary structure prediction). MBO is not a very good test as apomyoglobin
not is stable in water.
It is interesting that such a simple method seems to work that well.
Unfortunately this is one of many studies that would be extremely
much better if they had done some more rigourous testing, from
what I read in the John Hopkins journal, the run takes over night
or so, which makes me wonder why they do not run this on 50 different
proteins. I always get very very suspicious abot this method.
How do I know that they did not try it on 50 proteins and it only worked
in these 5 ? Probably not but they might have unintentionally have
optimised things for this test set. Any study that do not contain
a jack-knife test is much less impressive to me.
But atleast it gives me a few new ideas and I guess that makes
it a good paper. It is also described (more or less) good enough so
that anyone can implement the algorithm and try it themself.
arne
--------------------------------------------------------
From: Arne Elofsson
Email: arne at hodgkin.mbi.ucla.edu
WWW: http://www.doe-mbi.ucla.edu/arne/main.html
