AI-Assisted Molecular Discovery: Accelerating Drug Development Through Machine Learning Models

This Paper Finding new drug are a long and costly activity and pharmacokinetic properties [1]. Standard high-throughput experimental screening techniques are resource-intensive and time-consuming, and require specialized laboratory infrastructure, making them less accessible for smaller research institutions. We propose an AI-assisted molecular discovery framework that uses cheminformatics and machine learning to help with virtual screening in the early stages. RDKit works with molecular inputs, like SMILES strings or names of compounds, to get useful molecular descriptors. These descriptors are used to teach models how to guess how drug-like and bioactive something is. Model training and validation use publicly available datasets like ChEMBL, PubChem, and GuacaMol, as well as laboratory Minimum Inhibitory Concentration (MIC) datasets. The system also lets you see molecules in 2D and interactive 3D through a web-based interface. Models for machine learning are trained offline and then used to make predictions in real time. Experimental results show that ensemble-based methods accurately capture structure-property relationships and make reliable predictions, making it easy and cheap to prioritise possible drug candidates