In the realm of materials science, deciphering the behavior of multi-atom systems stands as a cornerstone yet formidable challenge. Lithium, for instance, holds a pivotal position in high-energy-density batteries. To propel the next generation of energy storage technologies forward, it is imperative to accurately predict the internal energy and interactions within lithium clusters. However, as the number of atoms grows, the complexity of system interactions escalates dramatically. While deep learning models exhibit immense potential, the lack of high-quality data and their 'black box' nature constrain their practical application. The team, headed by Professor Pan Feng from the School of New Materials at Peking University Shenzhen Graduate School, is committed to exploring graph-theoretical structural chemistry methods. This innovative approach translates the microstructure of materials into mathematical graph theory and topological models, offering a novel solution to this intricate problem.