A framework for prediction of hierarchical protein function based on PPI network and semantic similarity

Using PPI (protein-protein interaction) networks to predict protein function is a major research directions in genomics. However, due to the complexity and diversity of protein functions, traditional prediction methods using PPI networks often yield unsatisfactory results. To enhance the accuracy in predicting protein functions, a novel protein function prediction algorithm based on protein interactions was proposed in this paper, which combined graph metathesis with protein semantic similarity methods, and utilized FunCat's hierarchical function prediction based on PPI networks and semantic similarity. The proposed model was validated using Saccharomyces cerevisiae as an example, with data mainly obtained from the MIPS database. The experimental results showed that prediction precision reached 92.87% at level-1 and 78.23% at the most used level-3, outperforming similar prediction frameworks. Compared to traditional PPI-based prediction methods, this study improved the accuracy of protein function prediction and allowed for specification of the functional level of the predicted protein, resulting in directional prediction results.