RE: RE:Protein force fields



 Dear Chuck,
 As it happens I have just reviewed the Rao and Teeter paper for our journal
 club.  I chose the paper for 3 reasons: a) It was being presented to
 biochemists, b) I am a computational chemist, c) I liked the practicality
 of the approach.
 Martha Teeter is a crystalographer who has a long standing interest in proteins
 related to a-purothionin (eg crambin which she also solved), but has limited
 access to computing facilities (Microvax II).  She clearly wants to find a
 method that gives a good guess for solving protein structures by x-ray
 methods.  The paper presents a VERY rudimentary MD simulation using the
 AMBER UA force field, in vacuo, with much of the structure constrained.  This
 is a very simple and unsophysticated approach, and in some ways it is perhaps
 pleasantly surprising that it worked.  However, there are a few notable
 points:
 1) This is a purely anecdotal report.
 2) Alpha-PT is highly homologous to Crambin and so a very good guess for the
    MM/MD calculations was already available.
 3) The MD was performed on a very small subset of the atoms (6 out of 46
    residues only).
 4) Because of the particular nature of this protein (and its homologs) the
    ends of the critical loop-region could be assumed to be in the correct
    conformation, therefore extensive conformational restraints could be
    invoked.
 5) The authors report that an unconstrained MD simulation (epsilon=1,in vacuo)
    resulted in [hydrogen] bond breaking, but apparently did not test the
    stability with (epsilon=4) which they went on to use for the partial
    MD.
 6) Although the authors refer to a 2ps equilibration period, there PE
    data clearly shows equilibration was not reached until about 30 ps. This
    is unimportant in context of their analysis, but without continuing the
    trajectory for a longer period it is not possible to say how stable their
    loop conformation really is.
 The results are however interesting, particularly the sequential nature of
 the changes in the backbone conformation.  I think that if the system can be
 modelled based on similar high homology the UA approach looks very interesting,
 but it is very system dependant.  It worked in this case but I think solvation
 would have to be considered for more general studies.
 Hope this sparks other comments and answers some of your questions,
 Best wishes,
 Rob Woods
 Robert J. Woods, Ph.D.
 Glycobiology Institute
 Department of Biochemistry
 University of Oxford
 South Parks Road
 Oxford, OX1 3QU
 England