search

fraenkel-lab / OmicsIntegrator

This repository is the working directory for the Garnet-Forest bundle of python scripts for analyzing diverse forms of 'omic' data in a network context.
http://fraenkel.mit.edu/omicsintegrator
BSD 2-Clause "Simplified" License
31 stars 20 forks source link

msgsteiner directed edges #17

Closed agitter closed 8 years ago

agitter commented 8 years ago

@dylancronin and I found a dataset for which msgsteiner generates unexpected output. Specifically, the network it returns does not contain the DUMMY node, and several directed edges are used in the wrong direction. This causes Forest to throw an error when it parses the msgsteiner output because Forest does not find the incorrectly directed edges in its input list.

I don't think we need to change Forest's behavior because it correctly catches the invalid edges. I asked Alfredo for advice regarding msgsteiner.

agitter commented 8 years ago

Alfredo had a new version of msgsteiner, version 1.3, that fixed this problem. The code is available from http://staff.polito.it/alfredo.braunstein/code/msgsteiner-1.3.tgz

Before we recommend this as the default version for Omics Integrator, we should test whether it still passes our integration tests.

agitter commented 8 years ago

msgsteiner-1.3 passes the small test cases but fails the integration test: https://travis-ci.org/agitter/OmicsIntegrator/builds/149485021

I'll confirm that the new solution from 1.3 is better than the previous solution from version 1.1. Alfredo said that version 1.3 now returns the best solution found, whereas 1.1 erroneously returned the last solution found, so the new solution should be better.