Re: a bioinformatics question - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>Biology - 生物学

Re: a bioinformatics question

Re: a bioinformatics question# Biology - 生物学

w*u2004-06-08 07:06

1 楼

Agree. Currently, only homolgy modeling makes sense to biologists. The
state-of-the-art ab initio and fold recognition prediction methods are still
unable to reliablly predict even the correct fold (don't even think about the
full-atom, high resolution model which is very important for function
analysis). The problem here is how much you trust the predicted model before
you can go on to the next step for function analysis.

hypothesis is
free
structure

r*y2004-06-08 07:06

2 楼

I think it depends on the function of the protein. Some functions can be
easily predicted with its structure. For example, if it bind to some samll
ligands, then there should be a binding pocket. Or if binds with DNA/RNA(?),
then there should be highly charged part on the surface.. Protein-potein
binding sites are more difficult to be predicted.
There should be other info. to be used. If you get a protein structure right,
you have to assume a lot of things already. Such as if it's a membrane pro

i*e2004-06-08 07:06

3 楼

very wild question. Unless by homology modelling, no structures can be
reliablely predicted so far. Ab initio was even not recommended by CASP
committees last time.
I guess your question is that for a given structure without any homologous
structures solved, how to predict the function. If it is enzyme, you can try
to look for its largest or second largest pockets and do the sequence or
structure comparison upon the pockets. The reasonable assumption behind is
that structures are more conserved