Practical guidelines: Netter: re-ranking gene network inference predictions using structural network properties.

1. Running Netter

Netter uses a single command line argument: the path to a Netter configuration file. A typical execution of Netter is: java -jar netter.jar config_file.txt

2. A Netter configuration file

An example of a typical Netter configuration file:

# Use '#' to add comments 
# ---------------------------------------------------------
# Global parameter settings 
# ---------------------------------------------------------
testDir=data# folder location of the networks
preds=clr.txt# predictions that should be considered if allPreds is false, seperate with '='
allPreds=false # consider all the predictions in the folder, or only those specified in 'preds'
workersAvailable=2# amount of threads that should be spawned ( best <= amount of cores available on machine )
totalIterations=75# the total amount of iterations the algorithm should run
cutoff=300# the top X links of the ranking that should re-ordered
chuncks=50# the increase in size of subsequent subnetworks
chunckIncrease=0.5# the increase in penalty coefficient for subsequent subnetworks
modifyEachTurn=50# the maximum amount of modifications each turn (sampled uniform [1,modifyEachturn]
moveEachTurn=50# the maximum amount a link can move in the ranking in one modification, sampled uniform [-moveEachTurn, +moveEachTurn]
startTemperature=10.0# the starting temperature of the annealing schedule
maxIterations=500# the amount of turns of modification in one iteration
temperatureAutoDetermination=false# use an automatic scheme to set the temperature to a certain range
determinationZone=10.0# at which % of progress should we check if the temperature was set correctly ( default 0.1 =10%)
temperatureTargetGoal=0.12# the average target temperature of the automatic temperature scheme 
temperatureAllowedDeviation=10.0# the allowed deviation of the target goal without rebooting 
outputDirString=output# folder with output
# ---------------------------------------------------------
# ADD LOGGERS HERE, previous parameters should be set
# ---------------------------------------------------------
AcceptanceLogger=100=AcceptanceLogger.txt   # Specify classname, logging interval in iterations and outputfile name as arguments  seperated by = 
PenaltiesLogger=100=PenaltiesLogger.txt     # Specify classname, logging interval in iterations and outputfile name as arguments  seperated by =

# ---------------------------------------------------------
# ADD PENALTYFUNCTIONS HERE, previous parameters and loggers should be set
# ---------------------------------------------------------

The file can be split into four parts:
  1. Comments: All text following a '#" is ignored on a line.
  2. Global settings: All parameter settings of Netter not related to loggers or penalty functions. Description of each parameter is mentioned in comments.
  3. Logger to include: syntax 'ClassName'='Interval='OutputFileName' , the string 'LOGGERS=' specificies the start of the logger specification.
  4. Penalties to include: syntax 'ClassName'='Relative weight of penalty' , custom mapping of individual structure cost penalty can be specified by additional 'Key=Value' pairs ,the string 'PENALTIES= specificies the start of the logger specification.

3. Setting the parameters

Netter is shown to be robust for a wide setting of parameter values.

In practice, we suggest the following steps.

When using default penalty functions (basic): When using custom penalty functions (advanced):

4. Netter GUI

The Netter GUI is an interface to create Netter configuration files and track progress of running jobs. All computational functionality is implemented in the main Netter jar file.