Virtual Screening of Anticancer Drugs Using Deep Learning

In drug discovery, an efficient modelling of the synergies between the existing drugs/compounds and their targets is crucial. The application of In-vitro methods over millions of compounds is tedious and expensive. Virtual Screening, an In-Silico (Computational) technique has become as indispensable constituent of contemporary drug design. This technique executes efficient In-Silico searches over millions of compounds and drastically reduces the time and cost involved in drug design. This work intends to develop a Virtual Screening model for cancer drugs using Deep Learning. For this three Deep Learning algorithms were implemented on the dataset and their performance measures were recorded and compared to the performance measures given by the traditional Machine Learning algorithms on the same dataset. A significant gain in the performance metrics of the model was observed when the deep Learning algorithms were used. The activity of the molecules in the GDB13 dataset was predicted using this model to identify the potential anti cancer drugs.