SVM

1
Statistical Classification for Gene Analysis
based on Micro-array Data

• Fan Li Yiming Yang
• hustlf_at_cs.cmu.edu
• In collaboration with Judith Klein-Seetharaman

2
Principles of cDNA microarray
DNA clones
Laser 2
Treated sample
Laser 1
Reference
Excitation
Reverse transcription
PCR purification
Emission
Label with Fluorescent dyes
Robot printing
Hybridize target to microarray
Computer analysis
G. Gibson et al.
3
Microarray data how it looks like ?
Expression level of a gene across treatments
Expression matrix
Expression profiles of genes in a certain
condition
Typical examples Heat shock, G phase in cell
cycle, etc conditions Liver cancer patient,
normal person, etc samples
4
AML/ALL micro-array dataset

• This dataset can be downloaded from
  http//genome-www.standford.edu/clustering
• Maxtrix
• Each Row a gene
• Each column a patient (a sample)
• Each patient belong to one of two diseases
  types AML(acute myeloid leukemia) or ALL (acute
  lymph oblastic leukemia) disease
• The 72 patient samples are further divided into a
  training set(including 27 ALLs and 11 AMLs) and a
  test set(including 20 ALLs and 14 AMLs). The
  whole dataset is over 7129 probes from 6817 human
  genes.

5
Published work on AML/ALL

• Classification task gene expression -gt AML,
  ALL
• Techniques Support Vector Machings (SVM),
  Rocchio-style and logistic regression classifiers
• Main findings classifiers can get a better
  performance when using a small subset (8) of
  genes, instead of thousands
• Implication Many genes are irrelevant or
  redundant?

6
Possible Relationship (Hypothesis)
7
How can find such a structure?

• Find the most informative genes (primary ones)
• Statistical feature selection (brief)
• Find the genes related (or similar) to the
  primary ones
• Unsupervised clustering (detailed)
• based on statistical patterns of gene distributed
  over microarrays
• Bayes network for causal reasoning(future
  direction)

8
Possible Relationship (Hypothesis)
disease
9
Feature selection

• Feature selection
• Choose a small subset of input variable (a few
  instead of 7000 genes, for example)
• In text categorization
• Features words in documents
• Output variables subject categories of a
  document
• In protein classification
• Features amino acid motifs
• Output variables protein categories
• In genome micro-array data
• Features useful genes
• Output variables diseased or not of a patient

10
Feature selection on micro-array (ALM vs ALL)

• Golub-Slonim GS-ranking (filtering method)
• Ben-Dor TNoM-ranking (filtering method)
• Isabelle-Guyon Recursive SVM(Wrapper method)
• Selected 8 genes (out of 1000 in that dataset)
• Accuracy 100
• Our work (Fan Yiming) (best)
• Selected 3 genes (using Ridge regression)
• Accuracy 100

11
Feature selection experiments already done in
this micro-array data

• The 3 genes we found
• Id1882 CST3 Cystatin C(amyloid angiopathy and
  cerebral hemorrhage) M27891_at
• Id6201 INTERLEUKIN-8PRECURSOR Y00787_at
• Id4211 VIL2 Villin 2(ezrin) X51521_at

12
Some analysis on the result we get

• The first two genes are strongly correlated with
  each other.
• The third gene is very different from the first
  two genes.
• 1st gene 2nd gene is bad (10/34 errors)
• 1st gene 3rd gene is good (1/34 error)

13
QuestionAs the next step, Can we find more
gene-gene relationship?

• Several techniques available
• Clustering
• Bayesian network learning
• Independent component analysis

14
Clustering Analysis in micro-array data

• Clustering methods have already been widely used
  to find similar genes or common binding sites
  from micro-array data.
• A lot of different clustering algorithms
• Hierarchical clustering
• K-means
• SOM
• CAST

15
A example of hierarchical clustering
analysis(from Spellman et al.)
16
Our clustering experiment on AML/ALL dataset

• Our clustering result is over the top 1000 genes
  most relevant to the disease.

17
The feature-selection curve
18
Our clustering result in the top 1000 genes
19
Some analysis to the clustering result

• The first two genes are always clustered in the
  same cluster(in hierarchical clustering, they are
  in cluster 1. In k-means clustering, they are in
  cluster 2)
• The third gene is always not clustered in the
  same group with the first two genes(in
  hierarchical clustering, it is in cluster 23. In
  k-means clustering, it is in cluster 1)
• This validates our previous analysis.

20
Disadvantage of Clustering

• However
• It can not find out the internal relationship
  inside one cluster
• It can not find the relationship between
  clusters
• genes connected to each other may not be in the
  same cluster.
• Clustering vs Bayesian network learning(copied
  from David K,Gifford, Science, VOL293, Sept,2001)

21
A counter example of clustering analysis
22
Bayesian network learning

• Thus Bayesian network seems a much better
  technique if we want to model the relationship
  among genes.
• Researcher have done experiments and constructed
  bayesian networks from micro-array data.
• They found there are a few genes which have a lot
  of connections with other genes.
• They use prior biology knowledge to validate
  their learned edges(interactions between genes
  and found they are reasonable)

23
A example of the bayesian network

• Part of the bayesian network Nir Friedman
  constructed. There are total 800 genes(nodes) in
  the graph. These 800 genes are all cell-cycle
  regulated genes.

24
(No Transcript)
25
Our plan in genetic regulatory network
construction

• There are several possible ways
• Using feature selection technique to make the
  network learning task more robust and with less
  computational cost.
• Learning gene regulatory networks on microarray
  dataset with disease labels(thus we may find
  pathways relevant to specific disease).
• Using ICA to finding hidden variables(hidden
  layers) and check its consistency with bayes
  network learning result.

26
Our plan in genetic regulatory network
construction

• Use prior prior biology knowledge in gene network
  ,like the network motifs. The following example
  is copied from Shai S.Shen-Orr, Naturtics
  ,genetics, 2002. Previous network learning
  algorithm have not considered those characters.

27
(No Transcript)
28
Reference

• Using Bayesnetwork to analyze Expression Data ,
  Nir Friedman, M.Linial, I.Nachman, Journal of
  Computational Biology , 7601-620, 2000.
• Gene selection for cancer classification using
  support vector machines. Guyon,I.et al. Machine
  Learning,46,389-422.
• Clustering analysis and display of genome-wide
  expression patterns, Eisen,M.B. et al. PNAs,
  9514863-14868, 1998
• Clustering gene expression patterns . Ben-Dor,
  A.,Shamir,R., and Yakini,Z., Computational
  Biology, 6(3/4)281-297, 1999.