Information flow in interaction networks II: channels, path lengths and potentials

In our previous publication, a framework for information flow in interaction networks based on random walks with damping was formulated with two fundamental modes: emitting and absorbing. While many other network analysis methods based on random walks or equivalent notions have been developed before and after our earlier work, one can show that they can all be mapped to one of the two modes. In addition to these two fundamental modes, a major strength of our earlier formalism was its accommodation of context-specific directed information flow that yielded plausible and meaningful biological interpretation of protein functions and pathways. However, the directed flow from origins to destinations was induced via a potential function that was heuristic. Here, with a theoretically sound approach called the channel mode, we extend our earlier work for directed information flow. This is achieved by constructing a potential function facilitating a purely probabilistic interpretation of the channel mode. For each network node, the channel mode combines the solutions of emitting and absorbing modes in the same context, producing what we call a channel tensor. The entries of the channel tensor at each node can be interpreted as the amount of flow passing through that node from an origin to a destination. Similarly to our earlier model, the channel mode encompasses damping as a free parameter that controls the locality of information flow. Through examples involving the yeast pheromone response pathway, we illustrate the versatility and stability of our new framework.

💡 Research Summary

This paper extends a previously introduced random‑walk‑with‑damping framework for modeling information flow in interaction networks by adding a new “channel mode.” The original framework comprised two fundamental modes: an emitting mode that models how information spreads outward from source nodes, and an absorbing mode that captures how information is collected at sink nodes. While many later network‑analysis techniques can be mapped onto one of these two modes, they lack a principled way to represent directed flow from specific origins to specific destinations. The channel mode fills this gap by constructing a probabilistic potential function that combines the solutions of the emitting and absorbing modes for each pair of origin‑destination nodes. The core construct is the channel tensor, a three‑dimensional array whose entries quantify the amount of flow that passes through a given intermediate node when information travels from a particular source to a particular target.

Mathematically, the network is represented by a weighted adjacency matrix W and its associated Laplacian L. A damping parameter α∈(0,1) controls the probability that a random walker continues versus terminates, thereby regulating the locality of the flow. Solving the linear systems (I‑αW)h = source and (I‑αWᵀ)g = target yields the emitting and absorbing potentials h and g, respectively. The channel tensor C(i,j,k) is then defined as the product h_i·g_j·δ_{k∈path(i,j)}, where δ indicates whether node k lies on a path from i to j. This formulation provides a purely probabilistic interpretation of directed flow and allows the same damping parameter to tune the balance between short, local paths and longer, more global routes.

The authors demonstrate the utility of the channel mode on the yeast pheromone response pathway, a well‑studied signaling cascade involving receptors, G‑proteins, scaffold complexes, and MAP kinases. When the channel tensor is computed for all relevant source‑target pairs, key signaling proteins such as Ste2, Gpa1, Ste5, and Fus3 receive the highest flow values, matching biological expectations. Varying the damping parameter shows that the core pathway remains stable, while peripheral alternative routes become more or less prominent, illustrating the method’s robustness and its ability to capture multi‑scale network structure.

Compared with established techniques such as PageRank, HITS, and Random Walk with Restart, the channel mode uniquely quantifies directed, context‑specific flow without sacrificing the global probabilistic foundation of random walks. The tensor representation also opens avenues for downstream analyses, including functional module detection, disease‑gene impact assessment, and integration with heterogeneous or temporal networks. The paper concludes by highlighting future directions: extending the approach to multi‑layered biological networks, incorporating dynamic edge weights, and scaling the computation to large‑scale interactomes. Overall, the channel mode provides a theoretically sound, flexible, and biologically interpretable framework for directed information flow in complex interaction networks.