Preventing node ID's from changing?

[JaneeshBansal]

Hi,

I am running a simulation, where i take 2 haploid genomes and generate 4 recombinant products through 2 independent shuffling events. So there are mirrored products.

In the metadata section of the nodes table, I’m storing the node ID of the corresponding recombinant (mirror) product for a specific node. However, after running
tables.simplify(np.where(tables.nodes.time == 0)[0], keep_input_roots=True), the node IDs are reassigned.

Is there a way to either:

• preserve the node IDs through simplification
• automatically update the stored metadata with the new node IDs
• obtain a mapping between old and new node IDs after simplification

Thank you!
Janeesh

[hyanwong]

You can use filter_nodes=False to keep the IDs, or set map_nodes=True to obtain a node mapping.

There's a new tutorial on simplify() that I wrote about 3 days ago: https://tskit.dev/tutorials/simplification.html

[molpopgen]

You may want to consider that filter_nodes=False does result in a memory cost -- extinct nodes are retained in the node table.

Example to demonstrate:

import tskit

tables = tskit.TableCollection(100)

p0 = tables.nodes.add_row(0, 2)
p1 = tables.nodes.add_row(1, 1)
c0 = tables.nodes.add_row(tskit.NODE_IS_SAMPLE, 0)
c1 = tables.nodes.add_row(tskit.NODE_IS_SAMPLE, 0)

tables.edges.add_row(0, 100, p0, p1)
tables.edges.add_row(0, 100, p1, c0)
tables.edges.add_row(0, 100, p1, c1)

tables.sort()

ts = tables.tree_sequence()

print(ts.draw_text())

ts = ts.simplify(filter_nodes=False)

print(ts.draw_text())
print(ts.tables.nodes)

One strategy to save memory is to keep your metadata separate from the table collection until you are ready to make a final export to a file, do a final analysis, etc.. You use the "idmap" returned by simplify to modify your metadata as you go. When finished, you can use the "set columns" method on the node table to to write a final version that includes the columns already there + your metadata.

[petrelharp]

Another option here might be to, instead of saving "who is the partner to this node", save a unique ID in metadata that can be used to match them up later.