search

AuReMe / metage2metabo

From annotated genomes to metabolic screening in large scale microbiotas
https://metage2metabo.readthedocs.io
GNU Lesser General Public License v3.0
53 stars 7 forks source link

Non optimal powergraph visualisation #39

Open ArnaudBelcour opened 2 years ago

ArnaudBelcour commented 2 years ago

We have observed in several experiments that the visualisation offered by the powergraph does not reflect exactly the combinatorics of the solution. This seems to be normal as powergrasp uses some heuristics and can output sub-optimal visualisation.

But as it is also possible that powergrasp outputs optimal visualisation (such as the lipid powergraph in the metage2metabo article), we need to find a way to identify when a visualisation is optimal and when it is not.

A possibility could be to create automatically a list of equations using powernode names to list the generic associations of AND/OR of powernodes. Then using the nodes in each powernodes to compute the combinatorics. This would created the expected combinatorics from the powergraph visualisation. It could be compared to the observed combinatorics (stored in the community json in the enum_bacteria field). If there is a difference between the two then the powergraph is suboptimal. And if the two are equal then the powergraph is an optimal visualisation of the solution.

Also we have to add this in the readme and doc to avoid people misinterpretating the result of the powergraph.

ArnaudBelcour commented 10 months ago

Add a draft of a function to check this in commit 09f9a85944cef80ae11841b22db10d6530ba28a8.

ArnaudBelcour commented 9 months ago

Added a first draft in #61.