The invention belongs to the technical field of
viral vector design, and discloses a structural prediction-based
viral vector targeting optimization method and
system.The method comprises the following steps: firstly, obtaining structural information of a target
receptor and candidate
viral envelope protein; inputting structure information of a target
receptor into a trained
deep learning structure prediction model, outputting a
glycosylation shielding field associated with three-dimensional space coordinates, and quantizing reduction of space
accessibility caused by
glycosylation by numerical values of the
glycosylation shielding field; then docking the candidate
virus envelope
protein with a target
receptor structure, and predicting a candidate binding interface and a corresponding space coordinate set; extracting numerical values from the shielding
field based on the coordinate set, and obtaining a shielding punishment
score through predefined mathematical operation; constructing a multi-objective optimization function with other optimization objectives; and finally, iterating or screening candidate
virus envelope proteins by using an optimization
algorithm, and optimizing a function output result. According to the invention, the targeting of the
virus vector to target cells can be improved, and the off-target risk is reduced.