LogP estimation questions

[jle "at@at" world.std.com (Joe M Leonard)]

 Folks,
 	Kantola, Villar and Loew discuss an atom-based method for
 estimating Log P for molecules containing H, C, N, O, F and Cl.  The
 specific reference is "Atom Based Parameterization for a Conformationally
 Dependent Hydrophobic Index", A. Kantols, H. O. Villar and G. H. Loew,
 J. Comp. Chem. 12, 681(1991).  I have several questions about this work,
 and I hope somebody familiar with it is reading...
 1) It refers to Van der Waals radii from Gavezzotti, J. Am. Chem. Soc.
 105, 5220(1983).  Does anybody have these numbers on-line or at hand to
 post or email them?  The article implies that there are differences with
 these values as compared with "typical" values, and I am curious
 (havving
 spent some time collecting VdW radii).
 2) Has there been any corrections to this work?  If I compare my results
 against those published in the article, the results for F-containing
 compounds seem VERY different.
 3) Has this work been extended to additional atoms?  A later review
 article (Viswanadhan et al, J. Comp. Chem. 14, 1019(1993)) does not
 refer to this work.
 4) This work refers to the total charge of individual atoms.  In order
 to reproduce the published results, I have to substitute the valence
 charge (particularly for Cl).  This discrepancy, coupled with the
 fact that MNDO and AM1 are being used, suggests the valence charge was used
 without explicit reference.  Is this correct?
 I would appreciate any comments from folks interested in log-P estimation,
 particularly in regards to non-database methods (atom or fragment QSAR
 models, etc).
 Thanks in advance,
 Joe Leonard
 jle "at@at" world.std.com