Research Projects

The increasing number of multi-drug resistant disease variants and novel coronavirus pathogens pose significant challenges in human-healthcare. Wet-lab-based drug discovery projects are costly and time-consuming, making computational screening approaches less accurate. The main challenge is predicting binding free energies within a chemical accuracy of 1 kcal/mol. To manage healthcare problems, a reliable and fast scoring function is crucial. The current proposal aims at a machine learning and deep learning-based ranking method for identifying lead compounds. Another approach is to build a pharmacophore or develop a machine learning model for screening compounds from different chemical libraries. The Tanimoto index is used to compute the percentage similarity between compounds from different chemical libraries. Machine learning models can be classified or regression models to estimate binding affinity. The current proposal aims to apply recurrent neural network (RNN)-based approaches to generate novel lead compounds. The project also aims to identify strategies for parallel implementation of machine learning and deep learning tools to fully benefit from modern high performance computers (HPCs). Progress in parallel programming and multithreading libraries allows for the development of parallel and multithreaded versions of software for HPCs. Biological targets associated with neurodegenerative diseases such as Alzheimer's and Parkinson's will be considered in this project. Virtual screening of large chemical libraries using RNN will generate novel inhibitor molecules for these targets.