Multivariate design

Eriksson, L. and Johansson, E. (1996) Multivariate design and modeling in QSAR. Chemometr. Intell. Lab. Syst., 34, 1. 

If possible, investigate all of the biological models with a set of compounds with known clinical activity, designed in accordance with a multivariate design accounting for the clinical action as well as the chemistry. 

Kettaneh-Wold, N., MacGregor, J., Dayal, B. and Wold, S. (1994). Multivariate Design of Process Experiments (M-DOPE). Chemom.lntell.Lab.Syst., 23, 39-50. 

It is seen from the above points that the requirements on the experimental system for multivariate design are almost always fulfilled. 

The ability to recognize such interaction effects is therefore very important when a new reaction is elaborated, otherwise the potential of the reaction for preparative use may be overlooked. It is therefore necessary to use multivariate strategies to explore the reaction space so that the joint influence of varying the substrate, the reagent(s), and the solvent can be evaluated. This can be accomplished by multivariate designs in the principal properties. These designs will define sub-sets of test systems which can furnish the desired information. The principles are discussed in Chapter 16. 

The principles hitherto discussed are useful for "one-dimensional" studies of the reaction space, i.e. when variations along one axis only are considered. It was emphasized in Chapter 14 that interaction effects are to be expected when the solvent is varied with a simultaneous variation of substrate and/or reagent. To take such interaction effects into account it is necessary to use a multivariate design for selecting the test systems, so that the principal properties are jointly varied over the set of test systems. Examples of situations when this problem is encountered are ... 

The above example shows a selection of test systems by a design affording a uniform spread in the principal properties. The objective was to establish whether there was a gradual change in the performance of the reaction which could be related to the properties of the reaction system. The aim was to detennine those properties which have an influence on the selectivity so that these properties could be controlled. Both these objectives were attained. The results would have been very difficult to achieve without the PLS method and without using a multivariate design for selecting test items. 

Selected terms involved in multivariate design are as follows (15) ... 

Wold, S., Sjostrom, M., Andersson, P.M., Linusson, A., Edman, M., Lundstedt, T., Norden, B., Sandberg, M., and Uppg rd, L. Multivariate Design and Modelling in QSAR, Combinatorial Chemistry, and Bioinformatics. In Molecular Modeling and Prediction of Bioactivity. Plenum Press, in press, 1999. 

---