PLS loadings

We now have enough information to find our PLS Scores matrix and PLS Loadings matrix. First of all the PLS Loadings matrix is simply the right singular values matrix or the V matrix this matrix is referred to as the P matrix in principal components analysis and partial least squares terminology. The PLS Scores matrix is calculated as... 

The data matrix A x the PLS Loadings matrix V = PLS Scores matrix T (23-3)... 

Figure 33-1 presents the plot of the PLS loadings. Paul and Susan each computed both loadings. Note that the first loading is indistinguishable to the eye from the first PCA loading (see our original column on this topic [1]). 

Figure 33-1 PLS loadings from the synthetic data used to test the fit of models to nonlinearity, (see Colour Plate 2)...

Fig. 14.6 Correlation ofVolSurf descriptors with human intestinal absorption for 159 drug molecules. (A) Predicted versus experimental %HIA (human intestinal absorption) from final PLS model with 0.552, 0.709 and 4 PLS components. (B) PLS loadings for four-component...

The PLS scores are interpreted in the same way as PCA scores since they are the sample coordinates along the model components. The additional feature in PLS is that two different sets of components are considered, summarizing variations in the X space or Y space. PLS loadings express the relatedness of each X and Y variable to the model component. T scores are the coordinates of data points located in the X space that describe the part of structure in X which is most predictive for... 

It is worth noting that the optimal number of PLS loading vectors employed to predict various analytes in these papers is also consistent, ranging almost exclusively from 10 to 17, even for the largest of calibration sets. For integration times of a few minutes or less, it appears that any spectrum one acquires can be modeled (fit) to within the shot noise using no more than 20 independent spectral lineshapes. This then places a fundamental limit on the number of independent analyte concentrations that can be calculated from such data. 

The wa in equation (6) are the PLS loading weights. They are explained in the theory in references 53 - 62. Equation (7) shows how X is decomposed bilinearly (as in principal component analysis) with its own residual Epls A. T is the matrix with the score vectors as columns, P is the matrix having the PLS loadings as columns. Also the vectors of P and wa can be used to construct scatter plots. These can reveal the data structure of the variable space and relations between variables or groups of variables. Since PLS mainly looks for sources of variance, it is a very good dirty data technique. Random noise will not be decomposed into scores and loadings, and will be stored in the residual matrices (E and F), which contain only non-explained variance . 

Fig. 6. Correlation of VolSurf descriptors with human serum albumin binding affinity for 93 drug-like molecules. Left. Predicted versus experimental -log(K(HSA)) from the final 6 component PLS model. Right PLS loadings showing the importance of VolSurf descriptors to the prediction of human serum albumin binding.

Note diat in die implementation used in this text the PLS loadings are neither normalised nor orthogonal. There are several different PLS 1 algorithms, so it is useful to check exactly what method a particular package uses, although the resultant concen-tration estimates should be identical for each method (unless there is a problem with convergence in iterative approaches). 

Apply the sample, including any precipitate that may have formed, to an RNeasy mini column placed in a 2-mL collection tube. With the tube closed, centrifuge for 15 sec at 17,900g. Discard the flow-through. If the volume exceeds 700 pL, load and centrifuge aliquots successively onto the RNeasy column. After the RNA has been loaded onto the Qiagen column, it is ready for DNase treatment. 

In Equation (7.20a), pj and tj are, respectively, the PLS loading and score vectors, which are different from the loading and score vectors of the PCA method. In Equation (7.20b), and Uj are the loading and score vectors for the concentration matrix (C). 

---