CyclicCAE: A Conformational Autoencoder for Efficient Heterochiral Macrocyclic Conformational Sampling

CyclicCAE outperforms GenKIC by rapidly generating energetically stable heterochiral macrocycle backbones, advancing macrocycle drug design.

• Why it matters: Designing heterochiral macrocycles is challenging due to their complex conformations and lack of tailored computational tools, hindering therapeutic development.
• What they did: A convolutional autoencoder model was developed using a custom in silico dataset to efficiently sample macrocycle conformations and predict stable structures.
• The result: CyclicCAE produces stable, designable macrocycle backbones faster than existing methods, enabling energy minimization, diverse structure generation, and inpainting to accelerate drug discovery.