roy@phri.UUCP (Roy Smith) (09/26/87)
In article <2040@super.upenn.edu> li@linc.cis.upenn.edu (Siufai Li) writes: > Does anyone know if someone is doing research involving the prediction of > the shape of a protein from its amino acid sequence (or nucleotide sequence)? > [...] I wonder if such a task is possible using computers. This is without a doubt one of the big open issues of computational biology (does this field have an official name?) Lots of people are working on it, but as far as I know, nobody has had much success. The old standby is an algorithm described in: %A P.Y. Chou %A G.D. Fasman %T Prediction of the secondary structure of proteins from their amino acid sequence %J Adv. Enzymol. %V 47 %P 45-148 %D 1978 Chou-Fasman is almost a decade old and pretty much worthless as an analytical tool. Yet, because not much better has come along, people still use it (and, in my opinion, put far too much faith it the results). The ultimate goal is to take an amino acid sequence and predict not only the 3-dimensional structure, but also the function of the resultant enzyme, and its Km. Not that I think this ever will happen, but it's nice to dream. -- Roy Smith, {allegra,cmcl2,philabs}!phri!roy System Administrator, Public Health Research Institute 455 First Avenue, New York, NY 10016
dd@beta.UUCP (09/27/87)
In article <2910@phri.UUCP>, roy@phri.UUCP (Roy Smith) writes: > In article <2040@super.upenn.edu> li@linc.cis.upenn.edu (Siufai Li) writes: > > Does anyone know if someone is doing research involving the prediction of > > the shape of a protein from its amino acid sequence (or nucleotide sequence)? > This is without a doubt one of the big open issues of computational > biology (does this field have an official name?) Lots of people are > working on it, but as far as I know, nobody has had much success. See comments below. Depends on what you mean by "much success" > > Chou-Fasman is almost a decade old and pretty much worthless as an > analytical tool. Yet, because not much better has come along, people still > use it (and, in my opinion, put far too much faith it the results). This is an understatement; yet it remains the Gold Standard of structure prediction; if you come up with an algorithm, whether statistical, computational or heuristic (I'm thinking of the AI approaches of Abarbanel and company) you still have to compare your results to C&F. > The ultimate goal is to take an amino acid sequence and predict not > only the 3-dimensional structure, but also the function of the resultant > enzyme, and its Km. Not that I think this ever will happen, but it's nice > to dream. > -- > Roy Smith, {allegra,cmcl2,philabs}!phri!roy I think there has been more progress here than is immediately apparent, but only because the gains have been in limited domains. For example the Abarbanel (Biochemistry about 1983 and 1986) pattern recognition approach works very well...but on proteins which have alpha-beta-alpha structure. There was a paper by Stroud et al in either Immunology or the Journal of Immunology in which they use a power spectrum analysis to detect alpha helices which are amphipathic (specifically, hydrophobic on one side and hydrophylic on the other). Both these methods work with > 80% (I'm pretty sure it's actually higher) success, but cannot comment on joe-random- protein. The C&F approach suffers because its database was 29 globular, highly soluble proteins, and people have been applying the procedure to non-globular, amphi- pathic structures. The divide-and-conquer approach (pick a limited domain and get it right) has yielded better results but since you can't apply them to any protein they haven't made a splash. If anyone is interested I can post references to these and other articles. For instance, some time this year in PNAS there was an article comparing protein structure prediction methods for membrane-bound proteins. Regarding the full tertiary prediction prospects, yes we're a long way aways, but molecular dynamics and simulated annealing may help a great deal. Care to comment, dizzy Dan? dan davison / theoretical biology/ t-10 ms k710/lanl/los alamos,NM 87545 dd@lanl.gov ...cmcl2!lanl!dd