Diffusion Models at the Drug Discovery Frontier: A Review on Generating Small Molecules versus Therapeutic Peptides

Simple Summary Discovering new medicines is a slow, expensive, and often unsuccessful process. A new type of Artificial Intelligence (AI), known as diffusion models, shows great promise in changing this by designing entirely new drugs on a computer. This review examines how this technology is used to create two major types of medicines: small molecules, which are common in pills, and larger therapeutic peptides. For small molecules, the main challenge for the AI is to design drugs that can actually be created in a chemistry lab. For peptides, the focus is on designing molecules that are stable in the human body, fold into the correct shape to work properly, and do not cause an unwanted immune reaction. Both areas face common hurdles, such as the need for more real-world experimental data to train the AI and more reliable AI evaluation methods to predict a drug’s success. The ultimate goal is to connect these powerful AI design tools with automated robotic labs. This would create a rapid cycle of designing, building, and testing new medicines, transforming drug discovery from a process of slow exploration to one of creating novel, targeted therapies on demand.