Heat Diffusion

The heat diffusion algorithm implemented in PyBEL [2] is another approach that can be used to score and classify nodes in the candidate mechanistic subgraphs. Those nodes might represent an interesting biological process in the context of the disease or the pathology itself. The figure below summarizes the requisites of this algorithm:

  • A hypothesis graph: this might come from the output of the NeuroMMSig enrichment algorithm or using a subset of nodes of interest. The first workflow is already implemented in NeuroMMSig and the second it will also be there in the new version.

  • Differential data mappeable to the nodes: any kind of data that represent the difference between two groups (e.g., disease vs normal, group 1 vs group 2).

Figure: Heat diffusion algorithm. Summary of the requisites and real example of results.

With these two inputs, the algorithm will overlap the data into the nodes and run a simulation that spreads the heat over the network from sources to sinks. As a result, every node will have a final score (Fig 2. Table) that might serve, for example, to infer trends of those biological over time (Fig 2 Parallel coordinates).