Design of Highly Specific Structural Fragments for Filtering Compounds with Undesirable Activities

Abstract

The design of chemical compounds with desired properties is a key approach in medicinal chemistry for drug development. Modifying structural formulas it is possible to reduce undesirable biological activities exhibited by the substance under study. Various in silico methods are widely used to predict the types of activity in designed molecules, thereby reducing both the financial and time costs associated with experimental screening of potentially unsafe substances. However, the performance of many computer-based assessment algorithms depends on well-balanced training datasets containing a sufficient number of both positive and negative examples (information on the structural formulas of compounds that do and do not exhibit the desired type of activity). When standard predictive approaches are not applicable, an alternative strategy involves assessing the presence or absence of structural fragments associated with specific biological effects. To support this approach, we have developed a method that estimates the contribution of individual atoms to a given biological activity, taking into account their local structural environment. The applicability of this method to practical drug discovery tasks was demonstrated by analyzing molecular targets linked to a broad range of adverse drug reactions. Moreover, comparison of the constructed fragments with known structural motifs responsible for molecular target binding confirms the robustness of our method and highlights its potential for identifying functionally relevant structural regions in compounds that interact with molecular targets lacking resolved three-dimensional structures.