4.4 Discussion and Concluding Remarks

In the present study, we observed that the consideration of coexpression dynamics enables a strong localization power in PIN that reinforces module detection and characterizes phase-specific modules. We also found that the analysis centered on phases annotated in cores and communities reveals a certain convergence between the various PIN. Even if cores do not show the strong G1 characterization in cePIN that is observed with communities, the module annotation quality relatively to the matched complexes depends necessarily on the different resolutions allowed by the methods. Thus, the main driver of modularization is resolution-dependent, as expected.

The proposed approach of PIN fragmentation offers the possibility of looking at a compilation of PIN selected according to various criteria, for instance cell cycle specificity. An advantage is that comparative evaluations with regard to both general topological features and modularity can refer to multiple PIN referred to a common source. Thus, our study has referred to what we defined an “affine” PIN list, which opened the possibility to explore PIN dynamical aspects. Due to the consideration of time-course experiments and their recorded gene expression peak signatures, we could center the rest of the analysis on modularity with reference to the cell cycle role in determining the “interaction driver,” and the integrated gene measurements role in determining the “expression driver.”

Modularization has been mainly investigated relatively to clique-based methods. The comparison of community maps offered a coarse-grained analysis useful to verify what complexes are matched by modules and up to what extent, together with the involved pathways. Furthermore, a module characterization by phases can help monitoring the retrieved communities under different dynamic conditions. Then, the hierarchical fine-grained analysis obtained by comparing best and innermost k-cores was useful to point out the role that both module drivers may play at intramodular resolutions.

We have emphasized the major variation in the community maps by concentrating the analysis on hub proteins, thus reducing the dimensionality and complexity of modularity maps, and then validating at the protein pathway level the established community links. An increased phase localization power within the protein maps is observed when peak signatures are considered. Especially the modularization detected through k-cores tends to concentrate due to peak signature influence and intersection between both best and innermost hierarchical structures.

Two final notes concern a specific comment and a more general consideration. The modularization induced by the employed methods remains conditioned by the different resolutions that they allow to uncover. Depending on the network, differentiated modularity can be observed and maps are revealed according to the combined role of process-driven interactions and coexpression dynamics. In general, we believe that the development of differential network modularization approaches for examining cellular systems may be useful for inferring the complexities that typically characterize high-dimensional and heterogeneous biological data. In particular, the proposed approach for PIN could be extended to biological contexts where a crucial goal is establishing a role for biological processes involved in disease.