PLS ANALYSIS OF DISTANCE MATRICES TO DETECT NONLINEAR RELATIONSHIPS BETWEEN BIOLOGICAL POTENCY AND MOLECULAR-PROPERTIES

Although the statistical method of partial least squares (PLS) is wide ly used for the analysis of the relationship between molecular propert ies and biological potency, it is recognized that PLS detects only lin ear relationships. We tested two types of properties: simulated univar iate data and electrostatic molecular field as a function of Hammett a constants. In both cases we compared relationships in which the funct ion is linear, asymptotic, or rises to an optimum and then falls. We f ound that PLS analysis of the matrix of the distances between every pa ir of compounds detects all three types of relationships with the same quality of cross-validation. The successful application of the method requires that the distance matrices be constructed such that each con tains only information about one property (for example, the electrosta tic field around the functional group of interest). Carbo and Hodgkin similarities perform less well than distances.