The COVID-19 outbreak disrupted the world, highlighting the urgent need for efficient drug discovery. Traditional drug discovery methods, faced with an immense chemical search space for approximately \(10^{60}\) drug-like compounds, require substantial time and financial resources. Conventional in-lab techniques can test only \(10^{5}\) compounds per day, significantly increasing the cost and duration of drug discovery. Automating drug discovery through computational tools and algorithms is crucial to address these challenges. Drug producers increasingly integrate artificial intelligence (AI) techniques, such as graph neural networks and deep learning, to streamline drug discovery. Modeling molecular structures as string sequences has proven highly effective in solving numerous problems in drug discovery. Generative AI algorithms based on language modeling efficiently generate data by identifying patterns from training datasets. This study critically examines the algorithms and methodologies of generative AI in the context of drug discovery. It begins by exploring fundamental concepts and various generative models. The study then analyzed three prominent generative architectures: generative adversarial networks based on natural language, variational autoencoders, and generative AI models, evaluating their effectiveness in addressing key challenges in drug discovery. This study underscores the transformative role of generative artificial intelligence in advancing drug discovery, citing several prominent studies in the field. The review concludes by outlining future research directions and providing insight for leveraging generative AI models in the pharmaceutical sector.
