The present study establishes correlations of in vivo growth inhibition of a solid tumor, pancreatic ductal adenocarcinoma (Panc03), of mice with the steric and electrostatic fields and the hydrophobic parameter log P of a series (32) of 1-[[2-(dialkylamino)alkyl]amino]- 9H-thioxanthen-9-ones by the 3D-QSAR method comparative molecular field analysis (CoMFA). The template molecular model was hycanthone methanesulfonate (19), the structure of which had been established previously by X-ray crystallography. The hycanthone base is protonated at the terminal nitrogen N(2), and an intramolecular hydrogen bond is present between the proximal nitrogen N(1) and carbonyl oxygen O(1) atoms. Crystallographic data also indicate a planar arrangement of bonds around N(1). However, the molecular geometry of 19, optimized by semiempirical molecular orbital methods (PM3, MNDO, AM1), showed the expected trigonal-pyramidal configuration for N(1). A comparison of MO and ab initio methods applied to a model compound, 1-amino-9H-thioxanthen-9-one, led to the selection of PM3 as the method for full geometry optimization of first the cationic and then the neutral forms of 1-32, whereas AM1 provided atomic charges for these same structures save those incorporating a sulfonamide moiety (5, 7, 20, 25, 26, 29, 31, and 32). Acceptable values for the latter were obtained from ab initio calculations. Structures were aligned by minimizing root-mean-square (rms) differences in the fitting of structures to 19 using the FIT option of SYBYL. An alternative strategy of alignment, steric and electrostatic alignment (SEAL), was invoked to provide a comparison of statistical data generated with the rms alignment. The rms-fit alignment of structures produced slightly better cross-validated and conventional r2 values than those generated with the SEAL method. In addition, the rms-fit data indicate that a shift in the lattice of one-half of its spacing has a much smaller effect on the CoMFA data for a lattice of 1 A than one of 2 A. Inclusion of log P in a CoMFA of the neutral structures effected a small (ca. 8-10%) but significant improvement in cross-validated r2 values. The relative contributions of the hydrophobic effects and the steric and electrostatic fields to the conventional r2 values were 16%, 42%, and 42%, respectively. By contrast, incorporation of frontier molecular orbital (HOMO and LUMO) energies or their gaps in the PLS analyses failed to enhance correlation coefficients derived for either the charged or uncharged compounds.(ABSTRACT TRUNCATED AT 400 WORDS)