Cript Author Manuscript Author ManuscriptValdez et al.Pageand lacking the complement of immune cells present in stroma), it nonetheless delivers valuable data to illustrate the conceptual process of producing computational network models from IL-1RA Proteins custom synthesis dynamic profiles of paracrine signaling proteins, plus the relative physiological insights that will be discerned from utilizing data taken from the supernate measurement or the gel measurements. We analyzed the temporal protein concentrations obtained for 27 cytokines and development factors measured at 0, 8, and 24 hours post-IL-1 stimulation by constructing separate dynamic correlation networks (DCNs) for each in the two CC Chemokine Receptor Proteins Species information sets, i.e., these representing the external measurements (culture supernates) and those representing the nearby measurements (inside gels, by gel dissolution). Dynamic correlation networks are usually employed to infer transcriptional regulatory networks longitudinal microarray information. The approach computes partial correlations applying shrinkage estimation, and is consequently well suited for small sample high-dimensional data. In addition, by computing partial correlations and correcting for many hypothesis testing, DCNs limit the amount of indirect dependencies that seem inside the network and prevent the formation of “hairball” networks. Here, we use DCNs to identify dependencies among cytokines that may indicate either functional relationships or co-regulation. Considering that IL-1 is identified to trigger a variety of chemokines as well as other pro-inflammatory cytokines, which can further elicit signaling cascades (e.g. IL-6, TNF, MIPs and VEGF (60, 61)), we anticipated acute stimulation by exogenous IL-1 to correlate positively with (i.e., induce upregulation of) many in the measured cytokines while suppressing others. Within the DCN approach, relationships involving cytokines `nodes’ are elucidated by calculating correlation coefficients for each pair of cytokines/nodes across the 3 time-points (see Procedures), and then pruned to partial correlation partnership by removing indirect contributions amongst all potentially neighboring nodes. This DCN algorithm approach is specifically useful for acquiring trusted first-order approximations in the causal structure of high-dimensionality information sets comprising little samples and sparse networks (62). Fig. 5 shows the statistically substantial dynamic correlations, both optimistic and negative, comparing those identified for nearby in-gel measurements versus these found for measurements in the medium. From the nearby measurements, partial correlation evaluation discerns a highly interconnected cluster with two large branches stemming from IL-1 a single through MIP1 and an additional by way of IL-2. In contrast, precisely the same evaluation working with the measurements from the external medium does not connect these branches directly to IL-1 but instead confines its effect to a smaller set of associations, all of which are contained within the gel network. In addition to other differences that can be perceived by inspection of Fig. 5, this much more total network demonstrates that the local measurements far more fully capture the biological response anticipated from exposure to a potent inflammatory stimulus (IL-1) in comparison with measurements from the culture medium. Therefore, the neighborhood in-gel measurements might be a more correct strategy to reveal unknown interactions in complex 3D systems. These proofof-principle studies with cell lines demonstrate the potential for this approach for detailed hypothesis-driven mechanistic research with principal.
