Method for drug design using comparative molecular field analysis (CoMFA) extended for multi-mode/multi-species ligand binding and disposition

A computer-based method for generating a three-dimensional quantitative structure activity relationship of a series of ligand molecules in association with a common macromolecule, the method comprising:

• a. identifying one or more binding modes j for each ligand molecule i in the series of ligand molecules;
• b. placing each binding mode j for each ligand molecule i in said series into a grid for calculation of binding energies;
• c. determining in a multiplicity of grid points k for each binding mode j of each ligand molecule i, the interaction energy Xijk of binding mode j with a selected probe;
• d. expressing an association constant Ki for each ligand molecule i as a nonlinear function of the interaction energies Xijk for each binding mode j to yield a nonlinear binding function;
• and
• e. optimizing the regression coefficients in the nonlinear binding function;
• and
• f. calculating the optimized distribution of prevalencies of individual binding modes using the ratio of partial association constant Kij and the association constant Ki for each ligand molecule i with all the other ligand molecules in the series.

In a method for performing comparative molecular field analysis (CoMFA) of a three-dimensional quantitative structure activity relationship of a series of ligand molecules in association with a common macromolecule, the improvement comprising analyzing one or more binding modes j for each ligand molecule i in the series of ligand molecules by:

• a. identifying one or more binding modes j for each ligand molecule i in the series of ligand molecules;
• b. placing each binding mode j for each ligand molecule i in said series into a grid for calculation of binding energies;
• c. determining in a multiplicity of grid points k for each binding mode j of each ligand molecule i, the interaction energy Xijk of binding mode j with a selected probe;
• d. expressing an association constant Ki for each ligand molecule i as a nonlinear function of the interaction energies Xijk for each binding mode j to yield a nonlinear binding function;
• e. optimizing the regression coefficients in the nonlinear binding function;
• f. linearizing the nonlinear binding function to allow the use of partial least squares for iterative optimization of at least one regression coefficient to yield a linearized correlation function;
• g. applying a partial least squares procedure repetitively to the linearized correlation function until self-consistency to correlate the observed biological activity data with the interaction energies Xijk of the ligand molecules;
• and
• h. calculating the optimized distribution of prevalencies of individual binding modes using the ratio of partial association constant Kij and the association constant Ki for each ligand molecule i with all the other ligand molecules in the series.