While this is an interesting way to load data, it isn't standard, and doesn't scale well, since for each additional node involved you would have to add a new column and update all rows. Ideally, when the number of fields being set isn't growing, only the number of nodes needing to be connected or changed, your CSV should only need to change by adding additional rows. The number of columns should remain fixed.
A much easier approach is with a CSV that has a 3 columns: start, end, and value. Then you just build the CSV, each row representing a particular relationships between two nodes, and the value to use.
Your load query becomes much easier. Assuming that your nodes are already in the graph and this is just about creating relationships between them, and assuming we use :Node labels and :REL relationships (since that wasn't provided), and numeric ids for the nodes and numeric values for the values on relationships, a load of the relationships would be :
LOAD CSV WITH HEADERS from 'file:///myNewData.csv' as row
MATCH (start:Node {id:toInteger(row.start)}), (end:Node {id:toInteger(row.end)})
MERGE (start)-[:REL {value:toInteger(row.value)}]->(end)
For a sufficiently large CSV you would want to use USING PERIODIC COMMIT LOAD CSV instead.