A state-of-the-art software tool has been developed where all the successful multi-target QSAR models and in-house ADMET models were incorporated as a knowledge-base. The software has then been used to screen potential MTDLs against multiple G4s. The selectivity and binding characteristics of the screened MTDLs towards G4s over duplex DNA were consequently analysed by performing various in-silico and in-vitro assays.
The project tasks were:
⦁ Chemical and biological data curation of the collected experimental data. A comprehensive literature survey has been performed for identifying various ligand molecules along with their activity against various G4 motifs. Extensive data curation has been performed, including a complete checking & rectifying of errors in the chemical structure, exclusive handling of inorganic/organometallic/salts, normalization of the chemical structures, duplicate analysis, activity-cliff analysis etc
⦁ Multi-target QSAR models against different types of G4s have been developed. The type of models (regression and classification-based QSAR models) have been developed depending on the type of collected response data (continuous and categorical, respectively). The chemical numerical descriptors have been computed using available in-house python script and other freely available software. The list of descriptors included several classes such as constitutional, atom centered, connectivity indices, edge adjacency, electro-topological state, walk path counts, functional group, etc. As applicable, several linear and non-linear chemometric techniques have been employed to develop the models. Finally, the QSAR models have been evaluated using the standard protocol recommended by the OECD Guidelines.
⦁ An AI user-friendly, platform-independent software tool which utilizes the knowledge gained from the modeling study as well as the developed QSAR models to screen, optimize and/or design MTDLs against G4. This AI platform is based on KNIME nodes and KNIME workflow schemes.
⦁ A virtual screening campaign has been done using desirability-based multi-objective optimization, in silico and experimental evaluation of screened MTDLs. We have performed the virtual screening of big chemical space (databases such as, ZINC, Maybridge, DrugBank, InterBioScreen natural and Super Natural II, etc.), while employing desirability-based MOO approach. Screened ligands have been evaluated using molecular docking, MD simulations and key biophysical assays.
⦁ Events (conference, workshop, and other events) attended or conducted by the researcher: nine events described in the Tech. Report (Part B)
⦁ The software tool ‘G4-QuadScreen’ will be copyrighted. Our original intention was to also protect the models developed in the project, however, the hired agency to handle IPR informed us that there is no way to protect the models. However, since the models are implemented and automated in G4-QuadScreen, the original and validated models developed in this project will be available for end-users interested on them.
