Fig. 2
From: Graph embeddings on gene ontology annotations for protein–protein interaction prediction
Steps involved in our method of PPI prediction. Firstly, GO and GOA are combined together to construct an undirected and unweighted GOA graph. The node2vec model is then applied on the GOA graph to transform the nodes to their vector representations. Finally, the learned vectors are used for the tasks of missing and spurious PPI predictions. ti denotes a GO term and \({\mathbf {v}}_i = (v_{ij} )\) denotes its k-dimensional vector, \(p_m\) denotes a protein m and \({\mathbf {w}}_m = ( w_{mn} )\) denotes its k-dimensional vector representing protein. \({\mathbf {V}}_m\) denote a set of vectors of GO terms that annotate the protein