12.07.2015 Views

View - ResearchGate

View - ResearchGate

View - ResearchGate

SHOW MORE
SHOW LESS
  • No tags were found...

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Estimating Gene Function With LS-NMF 39with the first row giving the conditions (here, time-points) and the first columngiving gene names. The other data points then provide measurements of geneexpression for each gene in each condition.3.3. Setting Parameters and Running the SimulationLS-NMF requires four parameters for a simulation. The first parameter is theflag to enforce use of uncertainty estimates (see Note 4). The second parameteris the dimensionality of the factorization. As there is no available method yetto choose the dimensionality a priori, different values are usually tested in orderto find an optimal dimensionality for analysis, so the dimensionality parameteris usually a range. The third parameter provides the number of repeated LS-NMF simulations for each specific dimensionality (see Note 5). The fourthparameter is the maximum number of update steps, which terminates the simulationafter the specified number of steps whether or not the simulation hasreached convergence.These parameters are set using the interface of PattRun (see Note 6). At thispoint the PattRun algorithm should be started:1. On a Macintosh or Windows computer double-click on the PattRun.jar icon.This will launch the interface. On a Linux or Solaris computer, use the commandjava–jar PattRun.jar. The user interface will appear.2. Click on the Load button, and choose the file cdc25-sep1.txt using the file chooser.3. The interface will now appear as in Fig. 1, with the left hand column showing thedefault parameters for the simulation.4. Change the parameters for the LS-NMF simulation. Note the arrows in Fig. 2, andchange the numbers by clicking on the values and typing in the new values. UseuncertFlag 1StartRank 6EndRank 6Nchains 20which will set LS-NMF to use the uncertainty values in cdc25-sep1.unc, to tryonly six dimensions (herein to speed the process the correct number is identified),and to do 20 simulations. In addition, set the value of Niter further up in theinterface to 5000.5. The simulation can now be run, which is done simply by pressing the Run button.The display will note that the run is beginning, and it will update the status afterthe completion of each simulation.3.4. Interpreting the ResultsHow to interpret decomposed matrices from an NMF simulation depends onhow the original data matrix D is organized, and LS-NMF should be interpretedin the same way. Usually, matrix D provides the estimates of transcriptional

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!