Selwood dataset

It may be of interest to readers that all three methods mentioned above resulted in the same optimal subset of descriptors for the well-known Selwood dataset, which has become a de-facto standard in testing new approaches in this field [20]. 

MoQSAR has been demonstrated on a two-objective optimisation problem where the aim is to maximise model fitting, measured by r2, while simultaneously minimising model complexity, measured by the number of terms, p. When applied to the Selwood dataset with p allowed to vary between 1 and 7, a family of seven independent models (0.371complex models having a better fit to the data. The best four-term QSAR model found includes a cubic term (SURF A) and a quadratic term (SUM R) and is shown in Eq. (7). Note that whereas r2 has been optimised, q2 was calculated for the model with highest r2, i.e., it was not included in the optimisation process itself. 

Fig. 18. Scores corresponding to the first LV in the PLS model with the complete set of descriptors of the Selwood dataset.

Eig. 23. MEDA vector corresponding to the in vitro antifilarial activity (-LOGEC50) (a) comparison between this vector and that corresponding to ATCH6 (b) and MEDA vector corresponding to LOGP (c) in the PCA model with 10 PCs from the data set which combines -LOGEC50 with the complete set of descriptors of the Selwood dataset. 

Fig. 25. Plot of measured vs predicted values of -LOGEC50 in the model with regressors LOGP, ATCH6 and MOFI Y of the Selwood dataset.

In this section, an exploratory data analysis of the Selwood data set (34) is carried out. The data set was downloaded from http //michem.disat.unimib.it/chm/download/datasets.htm. It... 

---