Drug design - summary.

 Below is a summary of the replies to a question I recently posted.  I would
 like to thank everybody who replied to the question.
 First the question.
 > I am a math graduate student and working with one of the theoretical
 > the university here I developed a multipole algorithm for calculating
 > interactions.  With some modifications, this method can be used as a rapid
 > screening procedure for the electrostatic version of the docking problem.
 > The chemist that I am working with feels that this is a very important
 > in the area of drug design and that I should pursue it further.  In helping
 > to decide if I should pursue this, I was wondering if someone could briefly
 > summarize the state of the art in this field.
 The reply summary follows.
 > Hierarchical multpole methods are very usefull mathematical
 > tools in chemsitry.  See my page for papers with bibliography
 > on this topic.
 > On the other hand, attempting to fit or parameterize chemical
 > interactions with multipoles (or other functions) is an art, not a
 > science, as there are an infinite number of posibilities.  My two
 > cents is to focous on mathematical tools, and to avoid at all
 > costs anything that looks like a parameterization.
 > Cheers, Matt
 > +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 > Matt Challacombe
 > Los Alamos National Laboratory    http://www.t12.lanl.gov/~mchalla/
 > Theoretical Division              email: mchalla -8 at 8- t12.lanl.gov
 > Group T-12, Mail Stop B268        phone:   (505) 665-5905
 > Los Alamos, New Mexico  87545     fax:     (505) 665-3909
 I am somewhat familiar with the work of the above individual.  Two very
 interesting and novel papers of his are the following.
     1) Fast assembly of the Coulomb matrix: A quantum chemical tree code.
        J. Chem. Phys. 104(12) pages 4685-4698.
     2) Linear scaling computation of the Fock matrix.
        J. Chem. Phys. 106(13) pages 5526-5536.
 > The difficulty here, as I understand it, is that multipole methods
 > are useful only in the "far field" case, while it is nearby
 > that predominate in protein/ligand interactions.  I looked into
 > multipole methods for a slightly different, but related, purpose
 > some years back and decided that it wasn't worth it.  But don't let
 > my decision for a somewhat different problem and a less-than-perfect
 > understanding of multipole expansions discourage you from looking into
 > the problem more deeply yourself.  Just make sure that the part of the
 > calculation that you want to speed up is a sufficiently large part of
 > the total calculation that it makes a difference.  Remember that a
 > 90% speed-up of 10% of the calculation is only a 9% improvement overall.
 > 				regards,
 > 					Ethan A Merritt
 > -----------------------------------------------------------------
 > Dept of Biological Structure            K428b Health Sciences
 > University of Washington SM-20          (206)543-1421
 > Seattle, WA 98195-7742                  merritt -8 at 8- u.washington.edu
 Comment.  Multipole expansions have decent convergence only in the far field
 case.  However, I think this problem can be ameliorated with a couple of
 tricks.  The
 first is to fragment the molecules and compute the multipole moments for each
 the fragments.  The second trick would be to use non-linear convergence
 acceleration methods on the multipole expansion.  This idea has been tried on a
 spherical harmonic multipole expansion by H.H.H. Homeier in a paper recently
 published in the Internet Journal of Chemistry;
 http://www.ijc.com/articles/1998v1/28.  What I suspect to be
 the most difficult
 problem to work around are the induced effects.
 > Greengard-Rokhlin algorithm. Leslie Greengard is at Yale.
 > I don't know who has implemented this for finding molecular
 > energies. Use Science Citation Index to find out. Can
 > you  treat molecular models in solution, with one
 > dielectric constant inside the molecule, and a second,
 > in general different, dielectric constant outside the
 > molecule? I have a fast diffusion method for solving these problems.
 > We are using it to calculate solvation energies of macromolecules.
 >                            Best,
 >                            Jim Given
 >                            Center for Advanced Research in Biotechnology
 > While purely electrostatic interaction potentials were developed and
 > tried a few decades ago, they remain a very important part of
 > computational chemistry.  At present, most people do not use multipolar
 > representations for electrostatic interations (they stick with monomers
 > alone or use bond dipoles) since the computational time for multipolar
 > interaction calculations is significantly greater.  Don Williams
 > (Kentucky, I think) has some code which will fit electrostatic
 > potentials to a set of multipoles, of both atom and bond centered
 > character.  I believe that this code can also be used to compute and
 > export potentials using the derived multipoles.  There are a number of
 > other groups who have worked in this area, so please do not assume that
 > this posting is comprehensive.  Price, Stone and our own group
 > (Breneman) have also worked in this area.  Earlier workers are Ritchie
 > and Hirschfeld.  There is a rich literature in this area, but there is
 > always room for good ideas.
 > Prof Curt Breneman
 > RPI Chemistry Department
 > That work sounds very interesting and I would like to hear about any useful
 > replies you get please.  My colleague, Frank Burden (Chemistry Department
 > Monash University) have been working on molecular multipoles for screening
 > applications for some time.  We are basically improving the methods
 > developed by Silverman and Platt (Platt, D.E.; Silverman, B.D.  J.
 > Computat. Chem. (1996), 17, 358-66;
 > Silverman, B.D; Platt, D.R.   J. Med. Chem. (1996) 39, 2129-40).  The
 > critical question for drug design is how you define your axis system for
 > the electric multipoles with respect to the inertial axes.  If this can be
 > done correctly, the so-called 'alignment problem' (the need to superimpose
 > molecules which at act at the same receptor in a consistent way) can be
 > eliminated.  We do not agree with how S&P have defined theirs by feel
 > were on the right track. We have also added steric/inertial multipoles and
 > , in collaboration with Glen Kellogg at Virginia Commonwealth University,
 > have included 'hydropoles' (essentially expansions of the lipophilic
 > properties of drug molecules).
 > We would be interested in hearing more about your work.  Can you send us
 > any papers, reports, theses etc on what you have achieved?
 > Cheers,
 > Dave
 > Dr. David A. Winkler                    Email: dave.winkler -8 at 8-
 > Senior Principal Research Scientist     Voice: 61-3-9545-2477
 > CSIRO Molecular Science			Fax:   61-3-9545-2446
 > Private Bag 10,Clayton South MDC 3169   http://www.csiro.au
 > Australia 	        		http://www.molsci.csiro.au
 > Check out papers by
 > Greengard and Rokhlin,
 > S.Lustig and N.J.Wagner et.al.
 > sorry don't have them handy,
 > these are all quite recent publications, last maybe 4-5 years,
 > I am sure you'll find it in the database,
 > Hope this helps,
 > Mike
 > Michael J. Kotelyanskii	                     Phone (814) 863 43 81
 > Polymer Science Program			     FAX   (814) 865 29 17
 > Department of Materials Science and
 > Engineering                                  kotelyan -8 at 8-
 > Pennsylvania State University
 > University Park, PA 16802, USA
 > Yes, treating correctly Coulomb interactions in molecular simulations
 > (without using a cutoff in the list of interacting centers) is indeed an
 > important topic. The contest seems to have been won by smooth particle
 > mesh Ewald sums (SPME), which scales as O(N) like the fast multipole
 > technique, but with a much smaller costant factor, as I am said: see
 > T.A. Darden, D.M. York, L.G. Pedersen, J. Chem. Phys., 98, 10089 (1993);
 > U. Essmann, L. Perera, M. Berkowitz, T. Darden, H. Lee, L.G. Pedersen,
 > J. Chem. Phys., 103, 8577 (1995); P. Procacci and M. Marchi, J. Chem.
 > Phys., 104, 3003-3012 (1996); P. Procacci, T. Darden, M. Marchi,
 > J. Phys. Chem., 100, 10464-10468 (1996).
 > Two references for the fast multipole method, which I admit to have
 > never read though, are K.E. Schmidt, M.A. Lee, J.Stat.Phys. 1223-1235
 > (1991) and J. Shimada, H. Kaneko, T. Takada, J. Comp. Chem. 15, 28
 > (1994). See also the book D.Frenkel, B.Smit, Understanding Molecular
 > Simulation, Academic Press (1996).
 > It's not clear to me, however, if high accuracy and periodic boundary
 > conditions are needed in docking problems as well as in molecular
 > dynamics simulations. If they aren't, maybe the fast multipole method
 > which was originally devised for a cluster of ions is indeed a better
 > choice than SPME.
 > Regards
 > Dr. Guido Germano
 > Research Assistant in Theoretical Physics, University of Bristol, England
 > Tel. +44-117-928 8755, http://www.phy.bris.ac.uk/staff/germano_g.html
 > There is an abundance of reviews on drug design.  Look at the book series,
 > Reviews in Computational Chemistry, edited by Lipkowitz and myself.
 > Particularly Vol. 5 (1994) and Vol. 11 (1997).  These books will probably
 > in your chemistry department library.  Also, take a look at the May 1998
 > of CHEMTECH magazine, p.19.  Again it should be in your chemistry
 > library.
 > Don
 > Donald B. Boyd, Ph.D.
 > Editor, Journal of Molecular Graphics and Modelling
 > Department of Chemistry
 > Indiana University-Purdue University at Indianapolis
 > 402 North Blackford Street
 > Indianapolis, Indiana 46202-3274, U.S.A.
 > E-mail boyd -8 at 8- chem.iupui.edu
 > I suggest that you look at papers by S L Price (University College
 > London) to see applications of multipole calculations in molecular
 > modelling. However her work may not be directly applicable to
 > docking calculations: I think she is mainly interested in small-
 > molecule crystallographic modelling issues. Certainly multipole
 > calculations are an important technique in that area (although
 > they are rarely used in practice because the major modelling
 > software packages cannot handle them at present). I don't know
 > anything about docking myself so I cannot comment on how useful
 > they would be in docking calculations.
 > --
 >     John Osborn
 >     University of Bradford, UK.
 >     Email j.c.osborn -8 at 8- bradford.ac.uk
 Don Steiger
 dons -8 at 8- hamilton.math.missouri.edu