Analyzing cDNA microarray data using Python and the C clustering library: Why scripts are better than GUIs

1
Analyzing cDNA microarray data using Python and
the C clustering libraryWhy scripts are better
than GUIs

• Michiel de Hoon, Seiya Imoto, Satoru Miyano
• Human Genome Center, University of Tokyo

The third Bioinformatics Open Source Conference,
BOSC 2002 August 1-2, 2002, Edmonton, Canada
2
Scripting languages are already heavily used in
bioinformatics

• Bioperl http//www.bioperl.org/
• Biopython http//www.biopython.org/
• Bioruby http//www.bioruby.org/
• G-language http//www.g-language.org/ (uses
  Perl)

However, numerical analysis of cDNA microarray
data is still dominated by GUI-based codes
Why?
Because excellent GUI-based codes are available
for gene expression data analysis (such as
Cluster/TreeView by Michael Eisen, and
GeneCluster by Pablo Tamayo)
3
What scripting languages can do for you(Perl,
Python, Ruby, Tcl, )

• Easier to write, less prone to bugs, ideal for
  developing new algorithms
• Avoid checking your algorithm and chasing pointer
  errors at the same time
• More flexible than GUIs
• Allow batch processing
• Can run on any platform (Windows, Cygwin,
  Macintosh, Unix)
• Including supercomputers!
• Often compiler-independent (unlike GUI-based
  code)
• Makes open source software development easier
• A large number of people have contributed to
  scripting languages ?
• Software packages are often already
  available
• File handling
• Text parsing
• Graphics
• Numerics Data structures , algorithms, random
  number generators
• Beats writing a C/Fortran code from scratch
• One script can contain a complete data analysis
• Downloading data from a data base, file parsing,
  numerical data analysis, drawing figures
  invaluable for replication (see the example on
  our website)

4
Scripting languages make code development easier

• Write your new algorithm in Python
• Test it
• Improve the algorithm
• Implement the numerically intensive routines in
  C
• which can be called from Python
• thus combining the speed of C with the
  flexibility of Python
• If needed, the C routines can be used in other
  programs as well
• so scripting languages can make development
• of GUI-based codes easier too

repeat
5
An example The C clustering library

• The C clustering library contains routines for
  commonly used clustering methods
• hierarchical clustering pairwise single,
  maximum, centroid, and average linkage
• k-means clustering
• self-organizing maps on a 2D rectangular grid
• principal component analysis

• The C clustering library can be used in three
  ways
• by calling routines in the library from other
  programs
• as an extension module for Python
• through the improved version of the GUI-code
  Cluster/TreeView
• (which calls routines in the C clustering
  library)
• All three are available from our website.

The C clustering library as a Python extension
module has been compiled successfully on Windows,
Linux, and Unix (SGI-Cray Origin2000) systems
using GNUs gcc. No commercial compiler is needed
even for Cluster/TreeView. The library was
released under the GNU Lesser General Public
License.
6
At http//bonsai.ims.u-tokyo.ac.jp, follow the
link to
7
At our poster, you will find more examples of
using the C clustering library

• Downloading, analyzing, and clustering of gene
  expression data with Python
• An implementation in Python of a bootstrap
  calculation of hierarchical clustering
• Cluster/TreeView 3.0

How to find us Go to http//bonsai.ims.u-tokyo.ac
.jp, click on