Chemical space of orally active compounds

The aim of this work was to analyze chemical space of orally active compounds. A chemical space is a structurally diverse representation of a molecular database and contains regions of drugs having the same activity. Algorithms based on neural network systems or statistical multivariate data analyses have been used for clustering molecular data and mapping multidimensional information into lower level. We used a tree-structured self-organizing map (TS-SOM) technique to construct a chemical space of orally active compounds. In addition to the TS-SOM, other methods namely k-means and Sammon's mapping were used for clustering data. As a mapping result of the TS-SOM there are unrelated regions where well and poorly soluble and permeable drugs are mostly situated. Thus, it is possible to determine the regions where orally active drugs are located in respect to BCS, the biopharmaceutical classification system. The results of our study suggests that the properties related to the permeability and the solubility behaviour of drugs, such as molecular weight and the balance between hydrophobic and hydrophilic regions of the molecules, to be the most important feature in classifying the results of the TS-SOM. Moreover, the TS-SOM is proved to be an efficient tool for mapping high dimensional data into lower dimensions according to the properties of underlying data.