Abstract
Pharmaceutical industry is becoming more and more difficult in drug exploration and trial administration, as the expenses, long-term development, and more complex regulatory actions are rising. To respond to the mentioned challenges, the current research suggests a new model of drug discovery analytics that is constructed using a GAN, which is efficient in learning multidimensional patterns in molecules and enhancing predictive power. The data is preprocessed, features calculated by means of molecular descriptors, selected, and then the SMOTE is used to address unequal class distribution, and the model is then thoroughly trained and assessed. The suggested GAN model shows excellent performance, outperforming other ML and DL models with a (precision of 99.51), (recall of 99.99), (F1-score of 99.75) and (accuracy of 99.75). These findings imply that it has a high ability to generalize, less overfitting, and high robustness to classification tasks. On the whole, the framework provides an efficient, promising, and scalable idea in streamlining the drug discovery processes, facilitating better decisions, and speed-up innovation in pharmaceutical research and development.