Visualizing high dimensional descriptors of molecular structures

A method of visualizing in two dimensions the distance relationships in high-dimensional diversity space of compounds which are characterized by high-dimensional molecular structural descriptors comprising the steps of:

• a) selecting a representative subset of the compounds;
• b) evaluating the molecular structural descriptor characteristic of each compound of the subset selected;
• c) based upon the molecular structural descriptors, generating a distance matrix between every pair of compounds or utilizing a function to generate the distance matrix elements between every pair of compounds as needed;
• d) based upon the compound distance matrix, computing a hierarchy of clusters by defining cluster centers and partitioning the compounds at each level of clustering;
• e) performing a principal component analysis (PCA) on the molecular structural descriptors characteristic of each compound utilizing the first two PCA components;
• f) using a modified stress function which reflects a horizon, running a non-linear mapping (NLM) refinement from the initial PCA coordinates resulting from utilizing the first two PCA components;
• and
• g) graphically displaying the NLM generated coordinates of each compound as determined by the NLM refinement.