Towards the computational design of organic molecules with specified properties

In this work, we present and test a procedure for generating a chemical virtual library and its subsequent use to select molecular systems with desired properties. The library consists of molecular structures generated from a set of chemical fragments. As an example, we consider two tasks. The first one involves identifying structures with specific spectral properties, particularly concerning the UV–Vis region of the spectrum. To address this, the thiophene cycles with typical donor (dimethylamino) and acceptor (nitro) substituents are chosen as the molecular building blocks. First, the molecules from the derived virtual library are subject to computational screening using the semi-empirical tight binding density-functional method. The primary objective of the screening is to identify molecular structures that exhibit desired spectral properties, especially absorption in the long-wavelength region. Second, for the most promising structures identified in the initial screening, more accurate TD-DFT (B3LYP/cc-pVDZ) calculations are performed. Additionally, the advantage of the developed approach for library generation, aimed at further investigation of biological activity, is illustrated using an example involving papain-like protease (PLpro) inhibitors of the SARS-CoV-2 virus. The calculation scheme used in the considered examples is implemented in the Python program suite QUASAR.