Loading the GI50 of NCI60 into neo4j

load_gi50.cypher

// CSV file can be downloaded here:
// https://wiki.nci.nih.gov/download/attachments/147193864/GI50.zip?version=2&modificationDate=1649214698000&api=v2
LOAD CSV  WITH HEADERS FROM 'file:///GI50.csv' AS row 
MERGE (chem:Chemical {nsc: toInteger(row.NSC)})
MERGE (cell:CellLine {name: row.CELL_NAME})
MERGE (dis:Disease {name: row.PANEL_NAME})
WITH chem, cell, dis, row
MERGE (chem)-[:GI50 {concentration: row.AVERAGE, research: "NCI60", unit: row.CONCENTRATION_UNIT, experiment_id: row.EXPID, count: row.COUNT}]->(cell)
MERGE (cell)-[:CELL_LINE_OF]->(dis);

Sorry, it should be better out of the box, Imho in browser there are warnings for missing constraints/indexes.

The modeling question still stands though, i.e. would it make sense to model Experiment as a node.

@jexp Thank you for you help. I got it to work (need :auto before USING PERIODIC COMMIT in desktop version) this is extremely fast compared that what I was doing.

I was busy creating a benchmark between Redis and Neo4j where Neo4j is 15x slower with my stupid stupid query compared to redis. Is that noteworthy?
But with match normal queries they are the same

https://github.com/wagenrace/medical_data_blog/tree/bench_mark/Adding_NCI60/bench_mark

I think creating experiments would make sense, I stayed away from these question till I have more experience to set it up clearer. I want to add graph datascience (similarity and community detection) first and then revisit the problem again with more experience

I don't think it makes sense to create that benchmark, perhaps as a description of a learning experience, but I guess you have better things to spend your time on. (we have docs/courses on graphacademy that explain these things)

You might have even used our data import tool http://data-importer.graphapp.io/

If you want to try it here is a model + csv file that can be loaded from the "..." top right.
https://drive.google.com/file/d/1EPFRGPhvDSoE9hKGa05-iPHq2fztBQxd/view?usp=sharing