Page 109 out of 225 pages in this book.
Tripos Bookshelf  >  QSAR  >  QSAR Theory  >  CoMFA Theory

4.3 CoMFA Theory

The idea underlying a Comparative Molecular Field Analysis (CoMFA) is that differences in a target property are often related to differences in the shapes of the non-covalent fields surrounding the tested molecules. To put the shape of a molecular field into a QSAR table, the magnitudes of its steric (Lennard-Jones) and electrostatic (Coulombic) fields are sampled at regular intervals throughout a defined region. While there are many adjustable parameters in CoMFA, certainly the most important is the relative alignment of the individual molecules when their fields are computed. Properly aligned molecules have a comparable conformation and a similar orientation in Cartesian space. The QSAR is then generated by a PLS analysis of the data contained in the MSS. The value of the resulting QSAR can be determined through the value of the crossvalidated r² (from now on referred to as q²) reported by the PLS. If acceptable, the CoMFA QSAR, re-derived in final, non-crossvalidated form, can most easily be manipulated using various graphics techniques. Otherwise the alignment of one or more molecules can be changed, or other parameters altered, and the analysis repeated. Once an acceptable QSAR has been derived, prediction of the target property value for a new molecule is particularly straightforward (see below).

From a user's point of view, the four major phases of a CoMFA are:

• Preparing for CoMFA
• Building Spreadsheet for CoMFA
• CoMFA PLS Runs
• Examination and Use of CoMFA Results

Background information is also available on the following topics:

• Field Fitting
• Predictions and Missing Descriptor Values
• Extrapolation

To make it easy to get started with CoMFA, the QSAR AUTOCOMFA command carries out the last three phases automatically. Before using the AUTOCOMFA command, a database must contain the molecules of interest, all with atomic charges and all aligned relative to one another in a way that seems sensible to you. You are asked to enter the values of a target property for each molecule in the database for the values of a target property for each of the molecules in this database. SYBYL then constructs a region around the molecules and steric and electrostatic CoMFA fields are calculated for each molecule. A PLS analysis is then carried out using five crossvalidation groups, and the q² reported. If the q² is greater than 0.4, appropriate contour maps are generated.

More details follow about each of the four phases in a CoMFA, including the options available. Because CoMFA is so different from other QSAR and modeling procedures, and at the same time so apparently powerful, we have tried to provide more advice than usual on how to apply the method and what to do if problems are encountered.