Abstract

   The protein-protein interaction (PPI) network offers a conceptual
   framework for better understanding the functional organization of the
   proteome. However, the intricacy of network complexity complicates
   comprehensive analysis. Here, we adopted a phylogenic grouping method
   combined with force-directed graph simulation to decompose the human
   PPI network in a multi-dimensional manner. This network model enabled
   us to associate the network topological properties with evolutionary
   and biological implications. First, we found that ancient proteins
   occupy the core of the network, whereas young proteins tend to reside
   on the periphery. Second, the presence of age homophily suggests a
   possible selection pressure may have acted on the duplication and
   divergence process during the PPI network evolution. Lastly, functional
   analysis revealed that each age group possesses high specificity of
   enriched biological processes and pathway engagements, which could
   correspond to their evolutionary roles in eukaryotic cells. More
   interestingly, the network landscape closely coincides with the
   subcellular localization of proteins. Together, these findings suggest
   the potential of using conceptual frameworks to mimic the true
   functional organization in a living cell.
     __________________________________________________________________

   Proteins are basic parts of molecular machines that usually work
   together to perform their biological functions in a living cell. For
   better understanding the underlying cellular architecture and
   functional organization of the proteome, the protein-protein
   interaction (PPI) network provides a conceptual framework that depicts
   a global map of protein interactions in a topological
   space[32]^1,[33]^2. This framework has proven useful in systematical
   analysis of collective dynamics[34]^3, functional
   inference[35]^4,[36]^5,[37]^6, module identification[38]^2,[39]^7,
   signaling pathway modeling[40]^8,[41]^9, and other clinical
   applications, such as biomarker findings, disease
   classification[42]^10,[43]^11, and tumor stratification[44]^12.

   In a typical PPI network, proteins and their physical interactions are
   usually symbolized as nodes and edges, respectively, in a mathematical
   graph representation that describes entity relationships in the
   topological space. Proteins often work together to carry out their
   molecular functions by forming complexes or to engage in biological
   processes by interacting with each other in various interconnected
   pathways. These behaviors could be captured in the network model to
   detect functional modularity and protein cooperativity via in-depth
   topological analysis[45]^13. However, the inherent complexity of the
   biological network, which usually involves thousands of molecular
   entities and relationships, could make the systematic analysis
   difficult[46]^2,[47]^14. For example, due to the multi-functionality
   nature of proteins, a protein can play different roles and engage in a
   variety of biological pathway, thus creating multiple connections to
   various interacting partners in different biological contexts. This
   intricacy could limit the module detection and functional inference to
   relatively small local regions and also hampers in-depth investigations
   on global collective properties of the PPI network, such as its
   hierarchical structure and scale-free property, both of which are
   wildly conjectured on the global scale but their origins and the
   development processes are still unclear[48]^13,[49]^15,[50]^16.

   In social science studies, a common way to decompose a social community
   is to classify its members into age groups, based on the general
   observation that people of different ages also differ in their social
   roles, values, and positions in the community, and may potentially
   exhibit different behaviors in response to a given
   event[51]^17,[52]^18,[53]^19. We propose that the same approach could
   be applied to the biological network analysis. Since the cellular
   network, just like the genome, developed through
   evolution[54]^16,[55]^20,[56]^21,[57]^22, the phylogenetic grouping
   technique could be utilized as a tool to decompose a PPI network.
   Phylogenetics suggests the evolutionary relationships among species and
   proteins. A typical approach to classify proteins by age is to search
   for orthologs for each protein in other sequenced genomes and
   subsequently, the proteins can be assigned to age categories (groups)
   by tracing the latest common ancestral origin of their orthologous
   groups across phylogeny[58]^23,[59]^24,[60]^25. In this study, we
   adopted the same strategy, and combined it with force-directed graph
   simulation in the topological space, to decompose the human PPI network
   in a multi-dimensional manner. This approach, which we called
   phylogenetic decomposition (phylo-decomposition), enabled us to
   associate the network topological properties with evolutionary and
   biological implications.

   Briefly, our work proceeded as follows: First, we addressed the
   question whether proteins at different ages would play different roles
   in the human PPI network. From our phylo-decomposed PPI network, we
   observed that the ancient proteins occupied the core of the network
   with high topological centrality. Next, we examined if age homophily, a
   typical pattern of interaction preference in social networks, also
   existed in the PPI network. By analyzing the temporal patterns of
   interaction preferences within and between age groups, we revealed the