EFFECTS OF VARIABLE SELECTION ON COMFA COEFFICIENT CONTOUR MAPS IN A SET OF TRIAZINES INHIBITING DHFR

An example of a CoMFA study is described with the aim to discuss one o f the major problems of this 3D QSAR method: lack of variable selectio n. It is shown that the use of nonrelevant energy parameters might pro duce CoMFA contour maps which poorly reflect the actual nature of the binding site and are in part statistical artefacts. The data set emplo yed in our analysis comprises triazine inhibitors of dihydrofolate red uctase (DHFR), isolated from chicken liver, which have already been th e object of a QSAR study by other authors. Since three-dimensional str uctures of triazine-DHFR complexes are known, it was possible not only to reduce ambiguities in the superimposition of the ligands, but also to compare the resulting CoMFA contour maps with the enzyme active si te.