GAN-y w chemii: jak sztuczna inteligencja tworzy cząsteczki?

Artificial Intelligence (AI), particularly generative deep learning models such as MolGAN, CellGAN, and ORGAN, is gaining increasing importance in chemistry and structural biology. These models enable the creation of novel chemical compounds and molecular structures (MolGAN, ORGAN) as well as realistic cellular images (CellGAN), opening new possibilities in drug design, phenotypic analysis, and molecular engineering. This article presents an overview of selected generative architectures, including fundamental models and their specialized variants, discussing their advantages and limitations. Attention is drawn to the risk of generating unrealistic, hard-to-synthesize, or toxic molecules due to learning from statistical correlations rather than explicit chemical rules. Therefore, the need to integrate AI models with experimental knowledge is emphasized, along with the development of validation mechanisms and ethical safeguards in the context of their practical application.