MEO Node ID Error

[ntalluri]

If a node is in the format [name]_[info], MEO ignores the [info] part and only uses the [name] during reconstructions.

This happened when doing parameter tuning egfr dataset with MEO, the node names were missing the “_HUMAN” suffix. In the gold standard dataset, nodes such as EGF_HUMAN and EDGR_HUMAN are expected, but in the MEO output, they appear without the suffix. This error led to zero matching nodes to the gold standard nodes during evaluation.

input prize file

NODEID	prize	sources	targets	active
1433Z_HUMAN	1.041379133		True	True
41_HUMAN	3.389112802		True	True
4ET_HUMAN	2.569973509		True	True

raw pathway:

Source	Type	Target	Oriented	Weight
ERBB2	pd	CTNB1	true	0.553333
ERBB2	pd	CDK1	true	0.773333
EGFR	pd	SHC1	true	0.666667
EGFR	pd	41	true	0.553333

[agitter]

MEO may not support node IDs with underscores based on Neha's EGFR dataset testing and my initial inspection of its Vertex class: https://github.com/agitter/meo/blob/master/src/alg/Vertex.java#L45-L50

[ntalluri]

Doing something similar that domino does would help with fixing this issue

Ex.replacing the "" with a "-" and then in post processing, replacing the "-" with ""

[agitter]

The relevant DOMINO code is:

spras/spras/domino.py

Lines 215 to 231 in a26f4d0

	def pre_domino_id_transform(node_id):
	"""
	DOMINO requires module edges to have the 'ENSG0' string as a prefix for visualization.
	Prepend each node id with this ID_PREFIX.
	@param node_id: the node id to transform
	@return the node id with the prefix added
	"""
	return ID_PREFIX + node_id
	def post_domino_id_transform(node_id):
	"""
	Remove ID_PREFIX from the beginning of the node id if it is present.
	@param node_id: the node id to transform
	@return the node id without the prefix, if it was present, otherwise the original node id
	"""
	return node_id.removeprefix(ID_PREFIX)

I don't believe replacing "_" with "-" will work because then "A_B" and "A-B" would be the same. Those IDs are unlikely to both be in the network, but they could be. We'll need to replace "_" with another string that is very unlikely to appear in real protein names. If MEO supports Unicode, using some combination of strange Unicode characters would be safe.