LEAPFROG

[Richard M Day <r.m.day.,at,.cranfield.ac.uk>]

Dear Kathe, and the LEAPFROG user fraternity.
 Earlier this year I sent an EMAIL to the CCL, regarding the reliability of
 the LEAPFROG binding free energy calculations for ligand ranking and you
 were kind enough to email me with some answers.
 I have been using LEAPFROG for a while now to try and propose some xxxxxx
 binding ligands using oligopeptides as the initial approach.  Well to be
 completely honest none of the dozens of peptides proposed by LEAPFROG
 worked at all.  This is a problem for me,because my PhD was to use de novo
 and rational design to try and find something.  In the end I rationally
 designed a xxxxxxx with xxxxxx (censored) amino acids and this seems to
 work splendidly, phew...
 The key problem now is WHAT WENT WRONG ! and how could this be improved upon ?
 I have been trying to find literature where LEAPFROG has been evaluated,
 tried and tested by reproducing known ligands in binding site.  I have
 access to a very powerfull literature tool known as BIDS (based at Bath
 Uni, UK. www.bids.ac.uk) this has the abstract of every quality journal
 since 1980.
 We cannot find a single reference that refers to LEAPFROG.
 Talking to the UK technical support at Tripos, the algorithms in LEAPFROG
 are proprietary and therefore information on the precalculation of
 interaction points, empirical binding scoring mechanism, forcefields used
 if any, means of geometry optimisation and so forth, are not available for
 the world to see.  The only information I can find is a list of authors on
 the first page of LEAPFROG and some quite old references at the back.
 If LEAPFROG is based on the assumptions from these old references (after
 our  literature research, these assumptions are often have several flaws,
 of which many have been addressed in later versions) then this program is
 due some severe maintainance, of which we assume is already done since this
 is a proprietary program. But has it ?! If this has been done then why not
 tell us exactly what assumptions it uses ! and to top that, give us the
 references that they have based the algorithm so we can use both chemical
 insight with more understanding where and why the algorithms propose
 non-sensible solutions.
 There have been some quite fantastic developments in de novo design
 recently, including CONCEPTS, BRANCH&BOUND, PRO_LIGAND.  Why has not
 LEAPFROG adopted some new ideas ?
 I have heard from several sources that nobody is too bothered with de novo
 design today, but prefer VHTS with databases for initial lead finding. Even
 construct 3D stuctural databases with your own designer diversities.
 Databases of proprietary drug leads may contain upwards of 500,000
 compounds, but these have been sourced from complexes with active centers
 often in deep clefts within proteins. What happens if you want to look at
 active sites with only quite shallow binding clefts or even surface binding
 sites.  I cant imaging many compounds from such a database being a lead or
 "provoker" for such a surface borne active site.
 Well make your own database of structures and dock them !!
 -=-=-=-=-=-==-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 How are you supposed to design a diversity library for VHTS if you dont
 know where to take your "base" structures from ?  Answer, the use of a
 de
 novo tool.  The arguement that de novo design techniques are not worth
 developing because nobody wants them, is misguided.
 So to put it overly bluntly.
 Where is LEAPFROG 1999, to go with the maintenance fees ?
 Yours Faithfully,
 Richard Day.