Open clescoat opened 4 years ago
M1 tries to minimize the number of nodes in modules of size lower to --smallest, but it does not "force" the nodes to belong to larger communities if the algorithm finds they are better separated from their previous modules. We could have forced them to join other modules, but probably their quality would be lower.
In terms of the DMI challenge for which M1 was created, we knew that nodes in communities of size 1 and 2 would not be considered in the evaluation process, but we preferred to leave them as decided by the algorithm.
Summary M1 method doesn't take the --smallest parameter when running: even when the default setting is set explicitly (--smallest=3), clusters of 1 or 2 genes are found in the result file.
Example monet --input=./input/retinal_toxicity_BIANA.txt --output=/tmp/retinal_toxicity_test/ --smallest=3 --method=M1 --container=singularity
2020-06-30-164224M1result-modules__retinal_toxicity_BIANA.txt 2020-06-30-164224M1console-output__retinal_toxicity_BIANA.txt
OS/environment CentOS Linux release 7.6.1810 (Core) singularity version 3.5.2-1.1.el7
Version monet: commit 2333b993fc6a88abb32d72ec70eed7c59957a749 Date: Fri Mar 20 19:25:01 2020 +0100
Thank you for your help !