You are on page 1of 10

Biclustering of Expression Data

Clustering??
Euclidean Distance To find Pattern.

Biclustering solution
NP-hard problem. Since it needs ! ! time to find optimal (Perfect) solution

Evaluating Score
mean squared residue.

: variance of the set of elements

Algorithms.
0) Brute-Force Deletion and Addition.
: Delete/Add row/col as greedy with H value * Greedy methods : no guarantee to optimal Complexity = o( (n + m) nm )

Algorithms.
1) Single Node Deletion
: Delete row/col as largest variance. * Each time need to recalculate variance Complexity = o(nm)

Algorithms.
2) Multiple Node Deletion
: Delete row/col which has larger variance than specific value. * Need to select appropriate parameter Complexity = o(logn + logm)

Algorithms.
3) Node Addition
: From single cell add row/col without rasing H value. Sometime it gives more good result than deletion. * No guarantee to bicluster is maximal Complexity = o(nm)

Bicluster !
From Local To Global

You might also like