Data Reduction and Viewing for Virtual Library Design

Calculation of ADME predictions is now routine and often high throughput. However, unlike many published studies in which calculated properties are validated with experimental data for a diverse collection of molecules, for virtual small-molecule libraries, there are usually no physical data to validate the predictions. Often, the molecules that are the subject of calculation are dissimilar to those molecules used to develop prediction tools. As a result, one is usually looking for trends in the prediction as a function of the selection of specific diversity reagents around a common core or scaffold. Thus, it is important to consider predicted molecular properties with care and to interpret the results with the proper level of expectation. 

Data Reduction and Viewing for Virtual Library Design 

It has been proposed that a sparse matrix provides superior coverage to the selection of apparently optimal diversity reagents. However, the final collection of molecules that 

---

T5.W Data Reduction and Viewing for Virtual Library Design 407... 

---