biolink-model
biolink-model copied to clipboard
biolink
Similarity metrics based on CMAP: how to represent these in Biolink Model?
Is your feature request related to a problem? Please describe. @cbizon reported, on behalf of the Molecular Data KP, "the molpro team has some interesting similarity metrics based on CMAP. They compare gene expression profiles with added compound to either gene expressions from gene X KO or amplification. So then they get relations like chemical A has a similar effect on gene expression as knocking out gene B. Or chemical A has a similar effect on gene expression as enhancing gene Q. And on top of that, there is the ability to do all of those calculations for a given set of genes for the gene expression (i.e. some kind of context for the similarity calculation)."
Describe the solution you'd like Need way to represent these similarity metrics in Biolink Model.
What working group (or team) did this request originate from? Molecular Data KP
Additional context "The actual biology is that (say) the chemical has a similar effect in gene expression as increasing or decreasing another gene. But the increasing/decreasing is important, so just saying chemical-similarto-gene doesn't tell you enough. So how do you represent that? Is it qualifiers on similarity or new predicates, or using some kind of action nodes, or what?" (@cbizon)
Tag relevant members for discussion @vdancik, @remontoire-pac, @sandrine-m
Maybe the issue #473 is raising an additional point that issue #503 is not dealing with: the ability to compute similarity at the gene set level (dealing with gene lists). This discussion could be linked also to the discussion on the past relay about ability to make query at the gene list level.
@sandrine-m - we plan on covering this use case at the data modeling team meeting this week (Oct 28) if you have the time to join us? We also documented some possibilities that we'd like feedback on here: https://github.com/biolink/biolink-model/discussions/870