Melo E. B., Ferreira M. M. C., "SEVERAL LOGP IN A SAME QSAR STUDY? AN EXAMPLE WITH HIV-INTEGRASE INHIBITORS". Ribeirão Preto, SP, Brazil, 03-05/09/2007: 6th International Congress of Pharmaceutical Chemistry (CIFARP 2007), CD-ROM Cifarp 2007 (2007) QM 042. Poster QM-042.

QM 042 - SEVERAL LOGP IN A SAME QSAR STUDY? AND EXAMPLE WITH HIV-
INTEGRASE INHIBITORS.

EDUARDO B. DE MELO (PG)^(1,2); Márcia Miguel Castro Ferreira (PQ)⁽²⁾.

⁽¹⁾Farmácia/UNIOESTE; ⁽²⁾IQ/UNICAMP. marcia@iqm.unicamp.br.

Several methods of deriving LogP from molecular structure for using in  QSAR/QSPR  studies
are currently available. How to select the methodology most suitable for a study?  Or  still, is it
possible and admissible to  build a model with more than one LogP?  A study of  LogP applied
to  a  set  of  thirty-three  carboxamides   inhibitors  of  HIV-integrase  was  carried  out   using
fourteen  methodologies.   This  study  was   performed   employing   correlation   matrix   and
Hierarchical Cluster Analysis  (HCA).  The  majority  of  the  logP  values  had  presented high
intercorrelation (R > 0.8),  meaning that they provide mainly  the same information. The  three
methods that presented the lowest intercorrelation  (R < 0.6)  and were best correlated with the
biological activity data (pIC50):

ACDLogPCSLogPIALogPpIC50

ACDLogP10.5500.3610.469substructure approach  based on
atom contribution

CSLogP10.5410.412 whole molecule approach  based
 on topological and E-state indices

IALogP10.448whole molecule  approach  based
 on topological  indices and neural
 networks

These results indicated  that  the  three  methodologies,  although  having   the  same  physical
meaning in  a  QSAR  model,  could provide different  information  to  the model,  which is in
turn  related  to  the  specific  information  used   in  each  methodology  for  LogP calculation.
Classic  QSAR models,  applying  multiple  linear regressin method (MLR),  were  constructed
using the BuildQSAR program. Four models were obtained:

ModelLogPRSF

1ACD; CS; IA0.7550.66111.955

2ACD; CS0.7460.65917.616

3ACD; IA0.7360.67116.505

 4CS; IA0.6140.7818.487

Model (2) was chosen as the best one.  In conclusion, it has been shown that it is possible the
use of more than one type of LogP in a same QSAR model,  but attention must be paid to the
calculation method, its correlation with others and information extracted from it.
Financial Support: FAPESP/CNPQ.