The Non‐linear Iterative Partial Least Squares (NIPALS) algorithm is used in principal component analysis to decompose a data matrix into score vectors and eigenvectors (loading vectors) plus a residual matrix. NIPALS starts with some guessed starting vector. The principal components obtained by NIPALS depends on the starting vector; the first principal component could not always be computed. Wold has suggested a starting vector for NIPALS, but we have found that even if this starting vector is used, the first principal component cannot be obtained in all cases. The reason why such a situation occurs is explained by the power method. A simple modification of the original NIPALS procedure to avoid getting smaller eigenvalues is presented.

Structure-taste relationships for 25 acyclic and 20 cyclic carbosulfamates were investigated by means of pattern recognition using different graph theoretical invariants as molecular substituent descriptors. The SIMCA method was used to classify the compounds into sweet and nonsweet classes. All selected graph theoretical invariants that are related to the "rooted" vertex were found to give promising results. Using the weighted path numbers and self-returning walks for the rooted atom as descriptors of substituents, we found 87% of acyclic compounds were correctly classified. Using the atomic path numbers for the rooted atom as descriptors of substituents, we found 81% of cyclic compounds were correctly classified. These results are better than previously used shape and size substituent descriptors. It may be concluded that the graph theoretical descriptors have great potential in encoding structure components in structure-activity studies (SAR) studies.

Application of a new chemometric system, SPECTRE, to quantitative structureactivity relationship (QSAR) analysis in agricultural drug design has been studied. The SPECTRE system was employed to analyze experimental data by calculating a statistical predictive model using an evolution of the PLS (Partial Least Squares) regression method. This new modeling method, which does not need any a priori knowledge about the chemistry involved, is compared with multi-linear regression (MLR) analysis where the performance depends upon knowledge provided by the researcher. The SPECTRE system is shown to be able to produce similar or even superior results when compared with the conventional method.

No abstract

No abstract