Consensus molecular alignment based on generalized procrustes analysis.

One of the most serious problems in three-dimensional quantitative structure-activity relationship (3D-QSAR) studies is selection of an alignment rule for molecular super position of the compounds in the data set. In 3D-QSAR analyses of structure-activity data, a reference compound in a defined conformation is chosen, and all structures in the data set are aligned with the reference in a pairwise manner. In subsequent steps, conformation/alignment-dependent descriptors are computed for the compounds and compared to those of the reference. This approach gives much weight to the arbitrarily chosen reference molecule and can introduce a bias in the results. Here an alternative, and more general, approach to molecular alignment is presented that is based on Generalized Procrustes Analysis (GPA). The result is a consensus alignment that uses all molecules in the data set and avoids the bias introduced in the pairwise alignment strategy.