mapping-commons/sssom2neo

Convert SSSOM TSV to nodes and edges CSV files that can be ingested by neo4j-admin import.

To build:

mvn clean package

To run, assuming you have some mappings called mappings.sssom.tsv:

java -jar target/sssom2neo-1.0-SNAPSHOT.jar \
    --input mappings.sssom.tsv \
    --output-edges edges.csv \
    --output-nodes nodes.csv

You can also run over a directory containing lots of mappings files, like the OLS SSSOM dataset:

java -jar target/sssom2neo-1.0-SNAPSHOT.jar \
    --input ./mappings/ \
    --output-edges edges.csv \
    --output-nodes nodes.csv

Now you have two files, nodes.csv and edges.csv.

Let's load them into Neo4j! Assuming you already have Docker installed, we can do this quite easily. We will populate a new folder called neo with our neo4j database. First we use neo4j-admin to import the CSV:

docker run \
    -v $(pwd)/neo:/data \
    -v $(pwd)/nodes.csv:/mnt/nodes.csv \
    -v $(pwd)/edges.csv:/mnt/edges.csv \
    neo4j:4.4.20-community \
    neo4j-admin import --force --database=neo4j --array-delimiter="u+0000" --nodes=/mnt/nodes.csv --relationships=/mnt/edges.csv

If everything worked correctly, the neo folder should now contain a neo4j database populated with the SSSOM mappings from nodes.csv and edges.csv generated by the code in this repo. We can now start Neo4j:

docker run \
    -v $(pwd)/neo:/data \
    -p 7474:7474 \
    -p 7687:7687 \
    --env=NEO4J_AUTH=none \
    neo4j:4.4.20-community

Hit up http://localhost:7474 to go forth and cypher!

Examples

Get all mappings for a given subject

This query returns all mappings to/from MONDO:0005015 (diabetes mellitus). Note the syntax (a)<-[mapping]->(b) goes both ways, so both outgoing mappings (defined by MONDO) and incoming mappings (defined by other ontologies) are included in the results.

MATCH (a)<-[mapping]->(b) WHERE a.id="MONDO:0005015" RETURN *

Get all mappings for a given subject (transitive)

We can use an arbitrary level of depth, e.g. to search for mappings up to 3 levels deep:

MATCH (a)<-[mapping*0..3]->(b) WHERE a.id="MONDO:0005015" RETURN *

This result set includes transitive mappings e.g. MONDO:0005015-hasDbXref->UMLS:C0011849<-hasDbXref-ORDO:101952-hasDbXref->UMLS:C0011860.

Therefore UMLS:C0011860 (Type 2 diabetes mellitus) is included in the result set. Note that this is a more specific term than we started with! This is a limitation of the lacking semantics of hasDbXref, and a good example of why ontologies should use richer mapping metadata.