Cirino L. R., Cirino J. J. V., Ferreira M. M. C., “ACTIVITY/CYTOTOXICITY QSAR OF 4-AMINOMETHYLTHIOXANTHENONE AND 5-
AMINOMETHYLBENZOTHIOPYRANOINDAZOLE DERIVATIVES”. Düsseldorf, Germany, 27/08-01/09/2000: 13th European Symposium on Quantitative Structure-Activity Relationships: Rational Approaches to Drug Design, Book of Abstracts, P.63 (2000). Poster  P63. Section: New Developments in Chemometrics.

  Lucicleide Ribeiro Cirino, José Jair Vianna Cirino, and Márcia Miguel Castro Ferreira
Inst. de Química - Universidade Estadual de Campinas, Campinas SP, 13083-970 BRAZIL

4-Aminomethylthioxanthenone   and   5-Aminomethylbenzothiopyranoindazole  derivatives   are
antitumor agents against murine pancreatic adenocarcinoma  03 (Panc03). In vivo activities  and
in vitro  tests  were carried  out   by  Perni et. al. [J. Med. Chem. 1998, 41, 3645-3654.].  They
expressed in vivo tumor growth inhibition as a  %T/C value measured at the maximum tolerated
dose  for each compound dosed  intravenously. They also carried out  tests  to evaluate  in  vitro
cytotoxicity potency  against  Topoisomerase II  (P338) and  DNA  biding  via  intercalation  (all
Compounds  have shown  moderate  to strong DNA  biding  via  intercalation).
In this work,  log transformation of  in vitro cytotoxicity  (IC₅₀) and DNA  intercalation  potency
(EC₅₀) are investigated through QSAR models which were built using  electronic,  structural  and
topological  descriptors.   The  electronic  descriptors,  namely   E_homo   (energy  level  of  highest
occupied  molecular  orbital,  or  ionization  potential  according  to  Koopman's  theorem),  E_lumo
(energy level of the lowest  unoccupied  molecular  orbital), electronegativity  ((E_homo + E_lumo)/2),
hardness  ((E_homo - E_lumo)/2), dipole moment,  polarizability,  molecular  surface  area,  molecular
volume,   enthalpy  of  formation  (D H),   potential  energy   (Q_elet),   and  atomic  charges  were
calculated using  the quantum  chemical  package  Spartan  for  AM1  semiempirical  calculations.
The topological  and  electropological  descriptors  were  calculated  using  the  package  WHIM-
3D/QSAR.  The  models  were constructed  by means of  partial least squares  (PLS)  regression
using  a  small number  of selected  descriptors (5  in  each  case).

Finally,  the  activity  %T/C   is  studied  using   KNN  (K^th Nearest  Neighbor)  method   and  3
descriptors.  It  is possible to classify correctly  45 active and  10 non active  compounds  (%T/C
>42  is  not  active). [CNPQ, FAPESP]