An Association Analysis Approach to Biclustering

1
An Association Analysis Approach to Biclustering
Gaurav Pandey, Gowtham Atluri, Michael Steinbach,
Chad L. Myers and Vipin Kumar Department of
Computer Science and Engineering, University of
Minnesota
e-mail gaurav_at_cs.umn.edu website
http//www.cs.umn.edu/kumar/dmbio
APPROACH
MOTIVATION
RESULTS
Association patterns are biclusters!
Disadvantage
Advantages
Bicluster Group of objects showing similarity
over only a subset of the features in a data
set. Problem studied extensively for microarray
data for finding various type of
biclusters Finds more functionally enriched
groups of genes than hierarchical clustering
Prelic et al, 2006

• Exhaustive (and efficient) discovery of
  biclusters.
• Can discover small biclusters owing to bottom-up
  search procedure.

• Need to binarize or discretize the original
  real-valued data set which causes a loss of
  information Becquet et al, 2002 Creighton et
  al, 2003 McIntosh et al, 2007

Functional enrichment for small classes (1-30
members)
Functional enrichment for large classes (31-500
members)
Range Support An anti-monotonic support measure
for real-valued data!
Constant addition biclusters
Constant addition biclusters
Constant value biclusters
Constant row (column) biclusters
Madeira Oliveira, 2004

• Constraints imposed
• Consistency of expression values
• Same direction of expression
• These conditions satisfied over substantial
  number of conditions

CURRENT BICLUSTERING APPROACHES
Fraction of patterns (biclusters) enriched by
several groups of small classes at p-value 1x10-5
Fraction of class covered by patterns
(biclusters) among several groups of small
classes at p-value 1x10-5
Common Issues
Define an objective function/measure for
coherence of a bicluster
Non-exhaustive Heuristic search scheme doesnt
enumerate all biclusters satisfying the specified
condition Bias towards larger biclusters Objectiv
e function/measure satisfied early Non-overlappin
g biclusters (some)
Reorder rows and columnsfor globalminimum
Eliminate rows and columns for local minimum
Eliminate rows and columns from random seed

• Can be used within an Apriori-like framework
  Agrawal et al. 1994
• Implementation at http//www.cs.umn.edu/vk/gaurav/
  rap.

REFERENCES

• Computational Approaches for Protein Function
  Prediction A Survey, Gaurav Pandey, Vipin
  Kumar, Michael Steinbach, Technical Report
  06-028, October 2006, Department of Computer
  Science, University of Minnesota
• Association Analysis-based Transformations for
  Protein Interaction Networks A Function
  Prediction Case Study, Gaurav Pandey, Michael
  Steinbach, Rohit Gupta, Tushar Garg, Vipin Kumar,
  Proceedings of ACM KDD, pp 540-549, 2007
• Association Analysis for Real-valued Data
  Definitions and Application to Microarray Data,
  Gaurav Pandey, Gowtham Atluri, Michael Steinbach,
  Vipin Kumar, TR 08-007, Department of Computer
  Science, University of Minnesota, 2008
• H. Xiong, G. Pandey, M. Steinbach, and V. Kumar.
  Enhancing data analysis with noise removal. IEEE
  Transactions on Knowledge and Data Engineering,
  18(3)304319, 2006.
• H. Xiong, X. He, C. Ding, Y. Zhang, V. Kumar, and
  S. R. Holbrook. Identification of functional
  modules in protein complexes via hyperclique
  pattern discovery. In Proc. Pacific Symposium on
  Biocomputing (PSB), pages 221232, 2005.
• H. Xiong, P.-N. Tan, and V. Kumar. Hyperclique
  pattern discovery. Data Min. Knowl. Discov.,
  13(2)219242, 2006.

Cheng Church (CC)
ISA
Coclustering
ACKNOWLEDGEMENTS
This work has been supported by NSF grants
CRI-0551551, IIS-0308264 and ITR-0325949.
Computational resources for this work were
provided by MSI.