Reconstruction of dynamic protein-protein interaction network via graph convolutional network☆ ☆

Traditional researches predominantly rely on static protein network models, which present highly averaged and idealized representations of protein networks. However, these models fail to accurately capture the dynamic character of protein networks, which vary over varied time, location, and reaction conditions. In order to tackle the issue, we suggest a technique for convolving dynamic protein-protein interaction networks (PINs) through the utilization of knowledge graph reconstruction of relational graphs. To address the constraint of using only one dataset and to encompass the unique attributes of several datasets, four datasets were chosen from different databases to assess the viability of the strategy. The results demonstrate that this method surpasses existing methods in terms of dynamics. We found that a longer length of timestamps used in reconstructing dynamic PINs leads to a better prediction. In addition, we identified three hub genes (NUSAP1, SCG3, and CKAP2L) in dynamic PINs that were shown to be significantly associated with glioma prognosis. We conducted protein docking tests to validate the accuracy of the results. The findings confirmed that the proteins selected by using PreDPPI aligned with the anticipated experimental outcomes.