Deep-learning and interpretable machine-learning models are developed for electronic circular dichroism prediction, spectral-feature learning and chemical-structure analysis, improving the efficiency and interpretability of molecular spectroscopy. We also study molecular self-assembly, catalysed assembly, supramolecular polymerization and chirality transfer to construct molecular assemblies with defined topologies and functions. Representative publications: Decoupled peak property learning for efficient and interpretable electronic circular dichroism spectrum prediction, Nature Computational Science (2025); Computed ECD spectral data for over 10,000 chiral organic small molecules, Scientific Data (2025); What can molecular assembly learn from catalysed assembly in living organisms?, Chemical Society Reviews (2024); Molecular Face-Rotating Polyhedra: Chiral Cages Inspired by Mathematics, Accounts of Chemical Research (2024); Assembled molecular face-rotating polyhedra to transfer chirality from two to three dimensions, Nature Communications (2016).