Main documents for MBBC

• Joo, Y., Booth, J., Namkoong, Y., Casella, G. (2007) Model-Based Bayesian Clustering (MBBC), Technical Report, University of Florida
• Booth, J., Casella, G., Hobert, J. (2007) Clustering using objective functions and stochastic search, To appear Journal of Royal Statistical Society B.
• MBBC manual (Detailed description on options in MBBC)
• MBBC tutorial with screen shots.

Links for required pre-installations

• Ox
  Free download
  Ox homepage
    Ox installation guide (in pdf or html)
• R
  R download

Download MBBC

1. Download and Install MBBC for Windows

MBBC20zip.exe - A self-extractable file containing subdirectories and files below:

MBBC2.0/MBBC.exe

is the MBBC executable file.

MBBC2.0/readme.html

briefly explains input and output in MBBC graphic user interface. This file is also linked in help menu in MBBC. Further detailed explanation is given in Joo et al. (2007).

MBBC2.0/notes-lambda.pdf

explains how to estimate the smoothing parameters.

MBBC2.0/work

contains the R programs, Ox programs and .dll files. All files in this directory must be preserved to run MBBC correctly.

MBBC2.0/example

contains two example data sets; wound data (646 genes, 12 time points, 2 replications) and GC_example data (7640 genes, 4 time points, 1 replication)

MBBC2.0/output

will contain saved analysis results; MBBC report, clustered profiles, and simulation history.

• Download R, Ox, and Qt-based C++ source code (GPL and QPL), which is used to build the graphic user interface of MBBC 2.0.

MBBC20_source_zip.exe - A self-extractable file containing source codes

README-src.txt – Guide for installation of Qt and Dev C++ and compilation of MBBC source codes.

Example with Corneal wound healing data (Booth et. al. 2007)

• MBBC report
• Simulation results on convergence of stochastic search algorithm with different initial partitions

Computational Limitation

• MBBC runs only in Windows operating system.
• We do not recommend clustering more than 1000 genes with MBBC, because the optimization algorithm may take too long time before the convergence. If expressions of more than 1000 genes are measured, the user may reduce the number of genes with the filtering algorithm in MBBC.
• Without using the filtering algorithm, MBBC is tested for a simulated data set of 50000 genes, 30 time points, and 1 replication. We found that, although the optimization algorithm takes too long time, MBBC program runs without causing any memory overflow error in moderate pc environment (CPU:Pentium 2.6GHz, Memory:1.25Gb).
• If too many pictures of clustered profiles are included in MBBCreport.pdf, the creation process of the pdf file may cause a memory overflow error. Therefore, when the number of resulting clusters is greater than 1000, the results are saved in MBBCreport.html and png files, but not in the MBBCreport.pdf file.