Pasqualoto, K. F. M., Ferreira M. M. C., "RECEPTOR-DEPENDENT (RD) 3D-QSAR MODELS OF A SET OF ISONIAZID DERIVATIVES BOUND TO InhA, THE ENOYL-ACP REDUCTASE FROM M. tuberculosis, USING PLS REGRESSION AND GENETIC FUNCTION APPROXIMATION". Ribeirão Preto, SP, Brazil, 03-05/09/2007: 6th International Congress of Pharmaceutical Chemistry (CIFARP 2007), CD-ROM Cifarp 2007 (2007) QM 035. Poster QM-035.

QM 035 - RECEPTOR-DEPENDENT (RD) 3D-QSAR MODELS OF A SET OF
ISONIAZID DERIVATIVES BOUND TO InhA, THE ENOYL-ACP REDUCTASE
FROM M. tuberculosis, USING PLS REGRESSION AND GENETIC FUNCTION
APPROXIMATION

Kerly Fernanda Mesquita Pasqualoto (PQ); Márcia Miguel Castro Ferreira (PQ)

The State University of Campinas - UNICAMP

Introduction.   Enoyl-acp   reductase   is  a  key   regulatory  step   in  fatty  acid  elongation.
Biochemical  evidence  has  suggested  that  isoniazid,  a  first-line  drug  for the treatment  of
tuberculosis (TB), blocks the mycolic acids biosynthesis in M. tuberculosis. These fatty acids,
as  well  as  the key enzyme responsible for their elongation  (InhA),  are considered attractive
targets for the rational design of  new  anti-TB  agents.  Purpose.  Construction of  (RD)  3D-
QSAR models,  using  PLS  regression and genetic function approximation  (GFA)  formalism,
for  a  set  of  isoniazid  derivatives  bound  to  InhA.  Methodology.  Ligand-receptor  (L-R)
molecular dynamics  (MD)  simulations were carried out for a set of  37  hydrazines bound  to
InhA (PDB entry code 1zid) at 310 K (biological assay temperature). The hypothesized active
conformations resulting from a previously reported  receptor-independent  4D-QSAR analysis
were used  as  the molecular geometries for each ligand  in  this  structure-based  L-R  binding
research.  Four water solvent molecules that participate in  L-R interaction were maintained in
the active site during the MD calculations. The dependent variable is the reported MIC values
against M. tuberculosis var. bovis.  The independent variables (descriptors) are scaled energy
terms of a modified first-generation  AMBER  force  field  combined  with  a  hydration  shell
aqueous solvation model.  GFA and PLS  regression were employed as the fitting functions to
developed 3D-QSAR models, using the WOLF program (The Chem21 Group, Inc.). Results.
The best model  (N = 30)  presented the following statistical measures:  r²  = 0.77;  q²  = 0.61;
LOF = 0.58;  LSE = 0.29.  The  bound ligand solvation energy,  the sum  in  electrostatic and
hydrogen bonding energies of the unbound ligand,  the bending energy of  the unbound ligand,
the electrostatic intermolecular L-R energy,  and the change in hydrogen bonding energy upon
binding were found as the important energy contributions to the binding process. The external
validation (test set = 6 hidrazines) was 83.33%.  Conclusion.  The 3D-QSAR model (310 K)
has good internal  and  external predictability  and  may  be regarded as representative  of  the
binding process of ligands to InhA.