Reconnect nodes of degree zero in the graph

[malabikasen]

When we do cross validation, especially node-based, we delete edges and nodes from the network. This step makes these nodes disconnected, i.e., have degree zero. Clearly, this disconnection happens to TFs in node-based cross validation, i.e., for the TFs in the fold that we have left out. Genes (targets) can also become disconnected in this process.

We need to address this point in two steps:

(i) for every TF that is missing from the network, we add artificial edges from that TF to every other node in the network. We always perform this step.

(ii) for every (target) gene that is missing from the networks, we add artificial edges from every TF to that genes. We perform this step only if the configuration parameter is passed by the user.

[tmmurali]

Shouldn't line 129 of GAERunner.py loop over presentTFs rather than onlyTFs? @malabikasen @adyprat

[tmmurali]

I think I understand. The previous set of edge additions would have added artificial edges for the TFs that are not in present TFs, so it is now OK to add dummy edges from every TF in onlyTFs. Is that right @adyprat?