EdoardoGruppi/Deep_Understanding_of_AI_Based_Drug_Discovery

The present study is finalised to determine the most advanced models in the literature capable of producing new high-quality molecules starting from well-known datasets. The selection is carried out through a series of evaluation processes. At first, the output samples of each method are evaluated according to certain physico-chemical properties such as Quantitative Estimation of Drug-likeness (QED) and Synthetic Accessibility (SA). Then, in a successive step, the assessment also includes the predicted activity towards one target protein. The final aim of the project actually is to better understand whether and how the performance of each model varies when the typology of the target protein is changed. The modified code used to run the models is provided in the GitHub repo provided in description.

Python