Tree based modeling

Hong et al. [23] developed an approach to screen potential ER binding that included two rejection filters, a tree-based model, and three structural alerts to predict and prioritize ER ligands and applied this to over 58,000 chemicals. The training set included data from 232 structurally diverse chemicals whose RBA spanned a 106-fold range. They validated the models with a testing set of463 chemicals that were tested and showed ER activity. 

Figure 18.4 Tree-based model. The model ring index, logP, Jurs-PNSA-2, shadow-XY, and displays a series of yes/no (Y/N) rules to classify Jurs-RPCS. The squares represent the rules the chemicals into active (A) and inactive (I) circle represents the categoric results [23].

Shi, L.M., Xie, Q., Wu, J., Perkins, R., Walker, J.D., Branham, W. and Sheehan, D.M. (2002) Prediction of estrogen receptor binding for 58,000 chemicals using an integrated system of a tree-based model with structural alerts. Environ. Health Perspect., 110 (1), 29-36. 

Step 1. Performance of data structure analysis on real study data (untransformed and transformed) to reveal hidden structure, patterns, and relationships in the data set. This involves data visualization (graphing and fitting) and exploratory modeling (e.g., tree-based modeling). 

Riley, M. Tree-based modelling of segmental duration. n In Talking Machines Theories, Models and Designs, C. B. G Bailly and T. R. Sawallis, Eds. Elsevier Science Publishers, 1992, pp. 265-273. 

---