Mining Coherent Gene Clusters A Modified Approach

1
Mining Coherent Gene Clusters-A Modified
Approach

• Presented by
• Morshed Osmani

2
Original Paper

• Mining Coherent Gene Clusters from
  Gene-Sample-Time Microarray Data1
• By
• Daxin Jiang
• Jian Pei
• Murali Ramanathan
• Chun Tang
• Aidong Zhang

3
Outline

• Description of the Original Paper Process
• Other Related Works
• My Idea
• Further Improvement
• Suggestion and Comments

4
Original Paper- Problem Description.

• Given a set of n genes G-Set g1,g2,..,gn and
  a set of samples S-Set s1,s2,..,sl form a n
  x l matrix M mi,j where mi,j is the
  expression level of gene gi (1in), on sample sj
  (1jl).
• Each entry of M i.e. mi,j is a vector of T data
  points.
• Thus M can be viewed as M mi,jt where (1tT).

5
Problem Description Contd.

• We are interested in finding subset of those
  genes that are coherent on the subset of samples
  during the whole time series.
• This is essentially a subspace clustering.
• Coherent measurement is done by taking Pearsons
  correlation coefficient of two time series into
  account.

6
Coherence Measurement

• Given two vectors mi,j1t and mi,j2t of gene gi,
  the coherence is given by ?(mi,j1t,mi,j2t)
• A gene gi is coherent across a subset of samples
  S ? S-Set, if all pair of samples sj1, sj2 ?S,
  ?(mi,j1t,mi,j2t) gtd . Here d is the minimum
  coherence threshold.

7
Strategy

• Step1 Generate all the maximal coherent samples
  set for all the genes meeting the criteria
  (S?mins, ? gtd).
• Step2 Find the maximal coherent gene sets from
  for the produced sample sets in previous step.

8
Enumeration Tree

• Given a set of samples Ss1,s2,,sl, the power
  set (all possible combination) can be enumerated
  systematically using a set enumeration tree. An
  example is given below for set a,b,c,d.

9
Pruning Rules

• 3 pruning strategies are used in the form of one
  Lemma and two Pruning Rules.
• Lemma 3.1
• Pruning Rule 3.1
• Pruning Rule 3.2
• Explanation will be given on board.

10
Computing Maximal Coherent sample Set
11
Maximal Coherent Gene Clusters

• Can use the same technique used to find coherent
  sample sets
• But will make our algorithm exponential. 1000
  gene will require 21000 enumerations.
• Alternate solution use inverted list and again
  use the sample axis.

12
Inverted List
13
Maximal Coherent Gene Clusters - Algorithm
14
Results

• Applied algorithm on MS microarray data with
  ming15, mins3 and d0.8
• 21 coherent gene clusters reported.
• Have biological significance (involved in the
  same type of process)

15
Scalability
16
Related Works

• Biclustering2 measures the coherence between
  genes and conditions (samples or time series).
• TriCluster3 finds clusters along the 3 axes
  (gene, sample and time/space).
• Pattern based clustering4 finds subspace
  clusters using attributes of objects.
• Pattern based clustering also comes with quality
  measurements 5.

17
Important Features

• Some features may be noticed in the current
  paper.
• More emphasis on gene axes
• No measurements for assessing cluster quality.
• Search space is still very large. For l no of
  samples the search space is in the order of 2lX2l
  in the worst case.

18
My Idea

• Develop method for pruning large search space
• Measure the quality of the clusters
• Optional feature may be added produce top k
  clusters
• Implement the modified solution and get some
  simulation result
• Compare the modified algorithm with original
  papers algorithm

19
Current Work in Progress

• Implementing the original paper algorithm in C
• Looking for different quality measurement of a
  cluster

20
Future Direction

• Implement the current algorithm.
• Incorporate quality measurement in the algorithm
  and implement that
• Compare the result of those two implementations.
• Research for different pruning methods

21
Reference

• D. Jiang, J. Pei, M. Ramanathany, C. Tang, and A.
  Zhang. Mining coherent gene clusters from
  gene-sample-time microarray data. In 10th ACM
  SIGKDD Conference, 2004.
• Cheng Y. and Church G.M. Biclustering of
  expression data. Proceedings of the Eighth
  International Conference on Intelligent Systems
  for Molecular Biology (ISMB), 2000.
• Lizhuang Zhao and Mohammed J. Zaki. TRICLUSTER
  an effective algorithm for mining coherent
  clusters in 3D microarray data. Proceedings of
  the 2005 ACM SIGMOD international conference on
  Management of data, Baltimore, Maryland, 2005
• Wang H., et al. Clustering by Pattern Similarity
  in Large Data Sets. In SIGMOD 2002.
• D. Jiang , J. Pei and A. Zhang. " A General
  Approach to Mining Quality Pattern-based Clusters
  from Gene Expression Data". In Proceedings of the
  10th International Conference on Database Systems
  for Advanced Applications (DASFAA'05), April
  18-20, 2005, Beijing, China.
• S. C. Madeira and A. L. Oliveira. Biclustering
  algorithms for biological data analysis a
  survey. IEEE/ACM Transactions on Computational
  Biology and Bioinformatics, 1(1)24/45, 2004.