neo4j：Cypher查询中的可选“步骤”

Question

I am trying to find relations between nodes with optional but specific nodes/relationships in between (neo4j 2.0 M6).

In my data model, 'Gene' can be 'PARTOF' a 'Group. I have 'INTERACT' relationships between 'Gene'-'Gene', 'Gene'-'Group' and 'Group'-'Group' (red lines in model image).

I want to boil this down to all 'INTERACT' relationships between 'Gene': both direct (Gene-INTERACT-Gene) and via one or two 'Group' (Gene-PARTOF-Group-INTERACT-Gene) .

---

---

Of course this is easy with multiple Cypher queries:

# direct INTERACT
MATCH (g1:Gene)-[r:INTERACT]-(g2:Gene) RETURN g1, g2
# INTERACT via one Group
MATCH (g1:Gene)-[:PARTOF]-(gr:Group)-[r:INTERACT]-(g2:Gene) RETURN g1, g2
# INTERACT via two Group
MATCH (g1:Gene)-[:PARTOF]-(gr1:Group)-[r:INTERACT]-(gr2:Group)-[:PARTOF]-(g2:Gene)
RETURN g1, g2

But would it be possible to construct a single Cypher query that takes optional 'Group steps' in the path? So far I only used optional relationships and shortestPaths , but I have no idea if I can filter for one or two optional nodes in between two genes.

Answer 1

You can assign a depth between zero and one for each of the relationships you add to the path. Try something like

MATCH (g1:Gene)-[:PARTOF*0..1]-(gr1)-[:INTERACT]-(gr2)-[:PARTOF*0..1]-(g2:Gene) 
RETURN g1,g2

and to see what the matched paths actually look like, just return the whole path

MATCH p=(g1:Gene)-[:PARTOF*0..1]-(gr1)-[:INTERACT]-(gr2)-[:PARTOF*0..1]-(g2:Gene) 
RETURN p

There is a bit of a bother with declaring node labels for the optional parts of this pattern, however, so this query assumes that genes are not part of anything other than groups, and that they only interact with groups and other genes. If a gene can have [:PARTOF] to something else, then (gr1) will bind that something, and the query is no longer reliable. Simply adding a where clause like

WHERE gr1:Group AND gr2:Group

excludes the case where the optional parts are not matched, so that won't work (it'd be like your third query). I'm sure it can be solved but if your actual model is not much more complex than what you describe in your question, this should.

I took the liberty of interpreting your model in a console here , check it out to see if it does what you want.