The invention discloses a
protein function prediction method based on a graph neural network and a GO term hierarchical relationship and a related device. The method comprises the following steps: obtaining
protein sequence data,
protein PPI data corresponding to the protein sequence data and protein GO term data corresponding to the protein sequence data; performing
feature extraction processing based on the protein sequence data and the protein PPI data to obtain a
protein feature vector; performing term
feature extraction processing based on protein GO term data corresponding to the protein sequence data to obtain a term
feature vector corresponding to each piece of protein GO term data; and mapping the
protein feature vector and the term
feature vector corresponding to each piece of protein GO term data into the same
semantic space, and performing
protein function prediction
processing in the
semantic space. In the embodiment of the invention, better protein prediction performance is achieved, and compared with a traditional prediction method, the performance is greatly improved.