On October 21, 2024, The teams of Tingjun Hou and Changyu Xie from CPS-ZJU and Dongsheng Cao from Central South University published a paper entitled Leveraging Language Model for Advanced Multi-Property in Nature Machine Intelligence Molecular Optimization via Prompt Engineering, a method for multi-property molecular optimization using prompt-MOLOPT. The algorithm uses the training strategy of prompt learning to realize the application of zero sample learning and few sample learning in multi-property optimization, so that the model can effectively deal with multi-property optimization tasks even under the condition of single property data training. In addition, by transforming the input data from the traditional canonical SMILES string into a more detailed and targeted format that includes five parts: property label to be optimized, pharmacophore separator, pharmacophore, group separator to be optimized, and group to be optimized, the algorithm is able to preserve the integrity of the pharmacophore while properties are optimized. The experimental results show that Prompt-MolOpt is superior to the existing methods in multi-property optimization. This method is expected to improve the success rate of molecular optimization and ensure the integrity of pharmacophore during optimization, providing an efficient, accurate and reliable optimization tool for molecular design and drug development.