Abstract
In today's information-driven era, machine learning is revolutionizing medicinal chemistry, offering a paradigm shift from traditional, intuition-based, and often bias-prone methods to the prediction of chemical properties without prior knowledge of the basic principles governing drug function. This perspective highlights the growing importance of informatics in shaping the field of medicinal chemistry, particularly through the concept of the "informacophore". The informacophore refers to the minimal chemical structure, combined with computed molecular descriptors, fingerprints, and machine-learned representations of its structure, that are essential for a molecule to exhibit biological activity. Similar to a skeleton key unlocking multiple locks, the informacophore points to the molecular features that trigger biological responses. By identifying and optimizing informacophores through in-depth analysis of ultra-large datasets of potential lead compounds and automating standard parts in the development process, there will be a significant reduction in biased intuitive decisions, which may lead to systemic errors and a parallel acceleration of drug discovery processes.