UVE-PLS

Polanski, J. and Gieleciak, R. (2003) The comparative molecular surface analysis (CoMSA) withmodified uniformative variable elimination-PLS (UVE-PLS) method application to the steroids binding the aromatase enzyme. J. Chem. Inf. Comput. Sci., 43, 656—666. 

Scheme I Schematic representation of (a) PLS model (b) UVE-PLS model and (c) matrix B, containing regression coefficients calculated by the leave-one-out cross-validation procedure and their mean values, standard deviations, and stability.

Results of PLS, i.e. of the full spectrum method, are presented in Table 3. The RMSCV and RMSEP values are much higher than the analogous values observed for the SMLR or GA-MLR models, but one can hope that the PLS models are more stable e.g. when instrumental problems occur. Still, one can try to lower model complexity by extracting relevant information from the original spectra. This can be done, for instance, by using the UVE-PLS or... 

Table 3. RMSCV, RMSEP, number of latent variables (f) for the PLS and UVE-PLS models.

The complexity of the RCE-PLS models is always lower than the complexity of the PLS models, whereas RMSEP is lower for yl, y3 and yS, and higher for y2 and y4. The ypredicted versus yobscrvcd plots for the test set according to the PLS, UVE-PLS and RCE-PLS models are presented in Fig. 11. 

If the PLS, UVE-PLS or RCE-PLS models are applied to the test set slightly contamined with white noise (data set Xtn), they allow an acceptable prediction, thus giving evidence of their stability. RMSEP for Xtn, observed for... 

For data highly contaminated with noise also the difference between UVE-PLS and RCE-PLS approaches becomes more evident. For illustrative purposes, in Fig. 14 the spectra (centered) of test set contaminated with high noise (simulated as randn 0.01, i.e. ten times higher, than for Xtn), the relevant and irrelevant components, extracted by RCE-PLS for modelling y4, are presented. As one can easily notice, the majority of noisy variables are properly identified and removed from the original noisy spectra. 

Table 5. RMSEP for the test set contamined with white noise (Xtn) for the GA-MLR, PLS, UVE-PLS and RCE-PLS models.

FEROS and UVES spectra. This standard PL relation was derived for the LMC as a whole (< [Fe/H] >= —0.4) and 6(Mv) is, effectively, the correction to be applied to a universal PL relation as a function of metallicity. A positive 5(Mv) means fainter than the standard PL relation. 

When the number of noisy (noninformative) variables is too large, PLS models may also supply rather poor predictive performance. In order to overcome such a matter, a number of techniques for the elimination of noisy variables or the selection of useful predictors have been deployed, such as iterative stepwise elimination (ISE), iterative predictor weighting (IPW), uninformative variable elimination (UVE), and Martens uncertainty test (MUT) (Forina et ah, 2007). 

Fig. II The ypreju u-ii versus y hsi-rve,i plots for the test sets according to the PLS, UVE-...

The Uving character of the anionic polymerization of PL has enabled determination of the absolute rate constants of propagation on free macroions (fe ) and macroion pairs ), since the overaU propagation rate coefficient was shown to be the result of various forms of active species. For the polymerization of PL in CH2CI2 or in dimethylformamide (DMF), the reactivity of the macroion pairs was found to be almost independent of the initial monomer concentration, and weakly dependent on the temperature [33a,b] (cf. AHJ in Table 9.2). 

---