QSAR Modeling for Predicting AquaticToxicity of Chemicals

1.2 Regression QSAR modeling The QSAR model was developed to predict chemical toxicity (-log EC 50 mmol / L) using EL-based modeling method (DTB, DTF). The distribution of the selected descriptors used for regression modeling is shown in the radar chart (Figure 1). Using the smallest MSE standard in the set of training and verification, we decided the optimal structure and model parameters using 10 times CV. The MSE average (10 runs) and the internal verification and training data of the proposed QSAR model were 0.56, 0.11 and 0.709 0.941 (DTB) and 0.70, 0.14 and 0.661, 0.940 (DTF), respectively.

The regression method is used to build QSAR models in form of mathematical formulas. This equation uses an independent variable (descriptor) to describe the change in one or more dependent variables (usually active). The QSAR model can then be used to predict the activity of new molecules to screen large numbers of molecules with unknown activity. Multiple regression is the standard method of multivariate data analysis. Also called regular least squares regression (OLS). The regression method estimates the value of the regression coefficient by applying a least squares curve fitting method. In order to obtain reliable results, a data set that typically has as many data points (molecules) as independent variables (descriptors) is needed. Format used in regression expression

These computer simulation methods include pharmacophore model identification or quantitative structure activity correlation (QSAR) analysis. However, these methods can only be applied to "forward" problems which require calculation of physical, chemical and / or biological properties from the molecular structure of known active molecules. Another problem, also known as the "reverse direction" problem, is to identify the appropriate molecular structure taking into account some of the desired physicochemical or biological properties. The procedure to at least partially solve this problem is called de novo drug design procedure. These programs generate a new active structure suitable for the active site of the target protein by searching the entire chemical space. However, they must solve huge combinations and nonlinear structures - attribute related problems, exhaustive search is impractical

A method of selecting a chemically active chemical molecule by binding to a predetermined target protein