An Optimization Method for Drug Design Based on Molecular Features

Xuan Liu, Xiaoli Lin, Fengli Zhou

Abstract: Computer-aided drug design can reduce the cost and improve the efficiency of drug development. Due to the fact that the features of a molecule are closely related to its structure, the commonly used SMILES representation can only characterize the relative arrangement order of atoms to atoms in the horizontal direction, missing the relative positional relation between atoms in space. Therefore, this paper proposes an optimization generation model based on AIS encoding and drug-target interaction constraints. The model employs AIS coding for feature representation of drug molecules. The local structural information is further enhanced on the basis of obtaining global information by aggregating the information of neighboring atoms or fragments of each atom. To further validate the effect of the targeted drug generation model, the main protease 3CLpro was selected as the targeted protein. Experiments show that the proposed optimization method can generate the molecules with high affinity.

External IDs:dblp:conf/icic/LiuLZ24a