MDDR database, drug-likeness

When applied to the MDDR ° and CMC databases, this filter classified almost 70% of molecules as drug-like , while in the ACD database only 36% of molecules were found to be drug-like . The advantage of this filter is that it provides a detailed structural reason for classifying molecules as drugs or nondrugs. 

Several types of artificial neural networks have been used as pattern recognition engines. In general, neural networks are mathematical models that emulate some of the observed properties of biological nervous systems, in particular the ability to learn complex relationships. Examples of drug-like molecules were taken from the WDI, CMC or MDDR databases, while compounds from the ACD database exemplified nondrugs. Molecules were represented by a large number of descriptors, such as ISIS... 

While not convincing from a statishcal perspective, the results in this section are consistent with a trend high-activity molecules published in the past decade of medicinal chemistry literature are more likely to be found in the large, hydrophobic and poor solubility corner of chemical property space. These results are not consistent with, for example, cell-based [41] and median-based [42] partihoning of biologically active compounds however, such analyses were performed in the presence of inactive compounds selected from MDDR[41] or ACD [42], with quite probably unrelated chemotypes. ACD, the Available Chemicals Directory [43], and MDDR, the MDL Drug Data Report [43], are databases commonly used by the pharmaceuhcal industry. 

---