Nenad Vilendečić1, Milana Grbić1*, Miloš Radovanović2 and Dragan Matić1
1Faculty of Natural Sciences and Mathematics, University of Banja Luka
2Faculty of Sciences, University of Novi Sad
milana.grbic [at] pmf.unibl.org
Abstract
Proteins are recognized as essential functional elements in a wide range of cellular processes, where they typically act as components of protein complexes. In most network-based studies, protein complex identification has been mainly performed using static protein–protein interaction (PPI) networks. More recently, dynamic PPI networks have been employed for protein complex detection in order to account for temporal variability in interaction patterns between proteins and to enable time-aware analysis of protein complex formation.
In this study, the problem of detecting protein complexes in dynamic PPI networks is investigated. A dynamic PPI network is constructed by integrating a static PPI network with time-aware gene expression data, thereby enabling changes in protein–protein interactions to be represented across successive time points. Based on this representation, dynamic node embeddings are generated using a temporal network embedding approach. This approach encodes both structural and temporal properties of the network into a lower-dimensional vector space. The resulting embeddings are then used for the detection of dynamic communities within the network. These communities are interpreted as candidate protein complexes, which are subsequently subjected to a specially designed filtering procedure aimed at reducing noise and improving biological relevance. Finally, the filtered candidate complexes are compared against a set of reference protein complexes for evaluation purposes.
The proposed approach enables the analysis of temporally varying interaction patterns and provides a basis for protein complex identification in dynamic PPI networks. Preliminary observations indicate the potential advantages of incorporating temporal information compared to static approaches.
Keywords: protein-complex, protein-protein-interaction-(PPI)-network, dynamic-PPI-network, embedding, dynamic-community