Predictions of protein secondary structures based on amino acids are significant to collect information about protein features, their mechanisms such as enzyme’s catalytic function, biochemical reactions, replication of DNA, and so on. In this paper, we propose a new technique to predict the secondary structure of a protein using graph neural network. Firstly, fabricate a graph from the dataset using the primary sequence (amino acid). Every node in the graph is assigned a unique embedding using orthogonal encoding which is the feature of the model. Then iterate the whole graph in order, to sum up, the neighbor node’s information according to the GNN method. After the process, a well-known neural network model support vector machine (SVM) is implied in the graph and finds out the 8 states of the protein secondary structure. The simulation results illustrate the accuracy is 76.89% seems promising on the ccPDB 2.0 dataset. The comparative analysis shows the proposed method is significantly better than conventional methods.