<tool id="KMeans" name="KMeans cluster" version="1.0.0">
  <description>an aligned matrix</description>
  <command interpreter="sh">galaxyToolRunner.sh visualization.KMeans -i $input -k $K -1 $min -2 $max -o $output</command>
  <inputs>
    <param format="tabular" name="input" type="data" label="Aligned matrix" />
    <param name="K" type="integer" value="10" label="Number of clusters" />
    <param name="min" type="integer" value="1" label="Minimum column to use for clustering" />
    <param name="max" type="integer" value="-1" label="Maximum column to use for clustering (-1 to end)" />
  </inputs>
  <outputs>
    <data format="tabular" name="output" metadata="input" />
  </outputs>
  <tests>
  </tests>
  
  <help>
    
.. class:: warningmark

This tool requires tabular data in matrix2png format (with column AND row headers). For more information about the required format and usage instructions, see the matrix2png_ website.

.. _matrix2png: http://bioinformatics.ubc.ca/matrix2png/dataformat.html

.. class:: infomark

You can use the "Align values in a matrix" tool to create a matrix, then use this tool to cluster the matrix with k-means.

.. class:: infomark

**TIP:** You can use the **min** and **max** columns to cluster a large matrix based on a subset of the columns. For example, you could cluster a 4000x4000 matrix on columns 200-300 by setting min = 200 and max = 300. This will greatly increase the efficiency of distance calculations during the k-means EM, and also allows you to cluster based on specific regions, such as promoters or coding sequences.

-----

This tool will cluster the rows in an aligned matrix with KMeans_. The implementation builds upon the KMeansPlusPlusClusterer available in commons-math3_.

.. _KMeans: http://en.wikipedia.org/wiki/K-means_clustering

.. _commons-math3: http://commons.apache.org/math/

  </help>
</tool>