Hetnet prediction of candidate remyelinating compounds

Multiple sclerosis destroys the myelin coating neurons leading to disability. Here, we aim to find small molecules compounds that will promote remyelination. We base our search on our drug repurposing hetnet -- a network with multiple node and relationship types -- which provides a comprehensive framework for data integration. The hetnet is stored in a graph database called neo4j, which we use to compute network features.

Our approach to identify candidate compounds is unsupervised. We look for compounds that highly connected to myelination or oligodendrocyte differentiation. We specify the types of paths we are interested in. For example we look for paths where a Compound - targets - Gene - interacts - Gene - participates - Biological Process where the biological process is myelination.

See query.ipynb for the queries and data/queries.tsv for the results.

Queries for visualization

These queries should be run in the neo4j browser. They were chosen to produce illustrative and appropriately sized visualizations.

// Find paths between Marimastat and myelination
MATCH paths =
  (s1:Compound)-[:BINDS_CbG]-(n1)-[:INTERACTS_GiG]-()
               -[:PARTICIPATES_GpBP]-(t1:BiologicalProcess)
WHERE s1.identifier = 'DB00786' // Marimastat
  AND t1.identifier = 'GO:0042552' // myelination
  // Expressed in the central nervous system
  AND exists((:Anatomy {identifier: 'UBERON:0001017'})-[:EXPRESSES_AeG]-(n1))
RETURN paths
UNION ALL
MATCH paths = (s2:Compound)-[:BINDS_CbG]-(n1)-[:INTERACTS_GiG]-()
             -[:REGULATES_BPrG]-(t2:BiologicalProcess)
WHERE s2.identifier = 'DB00786' // Marimastat
  AND t2.identifier = 'GO:0048709' // oligodendrocyte differentiation
RETURN paths

Installation

This repository assumes a neo4j server containing the hetio-ind database is running on port 7474.

License

All original content in this repository is released under CC0 1.0.