NeuroMMSig - mechanistic networks
NeuroMMSig is a central repository for mechanistic knowledge in neurodegenerative disorders. As previously mentioned, over 200 subgraphs representing knowledge around mechanisms in three of the major neurodegeneration disorders are cataloged in NeuroMMSig. These networks are not static and can be queried through https://neurommsig.scai.fraunhofer.de/. Furthermore, algorithms can be applied to them in order to generate hypothesis as we demonstrated.
A user interface allows for the simple submission of data and the search for genes or proteins, SNPs, or imaging features derived from clinical experiments. The ranking algorithm implemented in NeuroMMSig will show to which candidate mechanisms the data can be associated. NeuroMMSig offers multiple functionalities enabling graph mining and reasoning over the graphs (e.g., graph algorithms, smart search, exporting options, knowledge provenance, active data visualization and Sankey diagram representations for pathway analysis). Besides, we have also included real gene expression data allowing users to infer what the (co-)expression of the nodes in a variety of highly relevant gene expression experiments are.
NeuroMMSig is already powered by several algorithms: enrichment ranking algorithm and story finder. The enrichment ranking algorithm calculates an enrichment score for the data-mapped to subgraphs. The enrichment ranking algorithm returns a list of subgraphs with their correspondent scores and metadata information. Moreover, the story finder algorithm allows the identification of upstream controller of a biological process of interest. More details of this algorithm can be found in in the NeuroMMsig intro page.
The so-called story finder algorithm is based on path mining algorithm and the polarity of BEL. While the effect of a single node over a neighbor node can be easily inferred via a direct edge, the complexity of this task grows exponentially when the path between the nodes gets longer and other nodes start getting involved in the equation.
This algorithm allows to discover key regulators that are involved in multiple paths from a set of nodes to a sink node that might represent a phenotypic feature. In the figure above, a Sankey diagram represents all paths from the gene targets of a drug to a biological process of interest in Parkinson’s disease. It can be observer how all paths pass through the BCL-2-BAX pathway which is involved in cell death/apoptosis.
Focusing on AD and PD, NeuroMMSig contains over 126 subgraphs in AD and over 76 in PD. It is noteworthy to mention that the listed mechanistic subgraphs include the specific mechanisms being investigated in the clinic by our partners (e.g., astroglial inflammation, mitochondrial dysfunction, neuroinflammation).