Regression Analysis

Regression analysis is a method used to study the correlation between argument X (such as molecular physical and chemical features) and dependent variable Y (such as activity of drug).

1. Partial least square (PLS) method

PLS could operate regression analysis for arguments and one or more dependent variable. PLS acquires better correlation with more dependent variables. Cross test is necessary for PLS analysis. We must drop one or more species during cross test, and predict activity of the species according to the model which is built on other compounds.

2. Artificial neuron network
3. Genetic algorithm (GA)

 

 

Parameters

The parameters of QSAR could illustrate that inter-interaction force between drug and receptor, drug delivery, metabolism, and other procedures. And then, we evaluate drug activity and relationship of above parameters. The parameters which are often used include hydrophobic parameters such as distribution coefficient and dispersion parameter, polarizability parameters such as molecular modeling refraction index, electronic parameters such as Hammett constant σ, field and conjugative effect parameter, charge transfer constant, dipole moment, coefficient operated by quantum chemical calculation, steric parameters which is obtained by liner free energy relationship or geometry parameter, indicator variable, and other parameters such as molecular weight, geometry parameter, conformation enthalpy, molecular connectivity index, and topological index.