Most descriptive compound MDC method

Hudson et al. [26] describes a method called the Most Descriptive Compound (MDC) method for selecting representative subsets and a sphere exclusion method for selecting sets of compounds that cover the available property space. The MDC method aims to select subsets that most effectively represent the compounds in the original collection. It operates by calculating a vector / of N elements where there are N compounds. For each compound, the other compounds are ranked in order of distance to it. The reciprocal of the rank of each compound n is then stored in vector position / . The process... 

Hudson et al. (67) describe two parameter-based methods for compound selection. The most descriptive compound (MDC) method is aimed at selecting compounds that represent the population as a whole. An information vector is accumulated from the ranked Euclidean distance of each compound in the data set to all others. The most descriptive compound is that with the largest information, which equates to the compound with the smallest overall distance to all other compounds. The next compound is chosen to give the greatest additional information and so on. The sphere-exclusion method used attempts to select compounds that most effectively cover the property space. A compound is selected, say the MDC, and all compounds are removed that are closer to it than a user-defined radius. The... 

---