Major TODO items:
1. Implement directed mode.
	The user would use a slider bar to find a high scoring cluster around a node of interest.  Moving the slider bar
to the right would expand the cluster, moving all the way to the left would reduce the cluster to the single node of interest.
To find the complex from a given node, the neighborhood of selected node should be searched for a high scoring node and
base complex on that high scoring node.  Need to be able to specify parameters on the fly.

2. Analysis option to evaluate results using a statistical model, as done in the MCODE paper.

3. "CLAST" feature to compare one set of complexes with another (in CASTA format) using various similarity measures
e.g. Dice, Jaccard, original one from MCODE, etc.


Minor TODO items:
1. Advanced options menu: Thread and progress bar for scoring and (finding+displaying) complexes

2. Deal better with parameter changes:
-If scoring parameters change, entire network must be rescored and find complexes must run again
-If only cluster finding parameters change, results dialog box must be recalculated (currently you must reload the network or
use the advanced options to force this to happen)
-When user changes the parameters and runs MCODE again and selects to create a new view for the complex, the views should
be labelled with the parameters that were used to find them (and possibly grouped under their own subnetwork).

3. Feature to save and load parameters (low priority)

4. Mark seed node by a node attribute that can be visualized (e.g. by node color)

5. Option to color by score in the menu (global options, would also have a visual attribute for a seed node). User
can do this now using visual styles, so this would be a convenience feature.

6. Convenience feature to save all complexes as sif or GML files.

7. Allow filter to filter by k-core - currently only filters by 2 core - k could be a parameter

8. Allow different node scoring functions - currently only MCODE, could have just clustering coefficient

9. Visualize functional information from GO - provide functional consistency score (number of functional terms/protein/complex)

10. Convenience option to turn off tooltips and some warning messages e.g. can't select nodes b/c no view created (advanced user mode)


Refactoring TODO items:
1. Have MCODE return Cluster objects and implement a Java network clustering framework where MCODE is only one
network clustering algorithm. Others are e.g. MCL, spectral clustering, etc.  At least 6 papers have been published in the past
year describing others.


