Download Making Sense of Complicated Microarray Data Part II - MGH-PGA

Making Sense of Complicated
Microarray Data
Part II
Gene Clustering and Data Analysis
Gabriel Eichler
Boston University
Some slides adapted from: MeV documentation slides
Why Cluster?

Clustering is a process by which you can
explore your data in an efficient manner.
 Visualization of data can help you review
the data quality.
 Assumption: Guilt by association – similar
gene expression patterns may indicate a
biological relationship.
Expression Vectors
Gene Expression Vectors encapsulate the
expression of a gene over a set of
experimental conditions or sample types.
Numeric Vector
-0.8
1.5
1.8
0.5 -0.4 -1.3 0.8
1.5
2
Line Graph
0
-2
Heatmap
-2
2
1
2
3
4
5
6
7
8
Expression Vectors As Points in ‘Expression Space’
G1
G2
G3
G4
G5
t1
t2
t3
-0.8
-0.4
-0.6
0.9
1.3
-0.3
-0.8
-0.8
1.2
0.9
-0.7
-0.7
-0.4
1.3
-0.6
Similar Expression
Experiment 3
Experiment 2
Experiment 1
Distance and Similarity
-the ability to calculate a distance (or similarity,
it’s inverse) between two expression vectors is
fundamental to clustering algorithms
-distance between vectors is the basis upon which
decisions are made when grouping similar patterns
of expression
-selection of a distance metric defines the concept
of distance
Distance: a measure of similarity between gene expression.
Exp 1
Exp 2
Gene A
x1A
x2A
Gene B
x1B
x2B
Exp 3
Exp 4
x3A
x3B
x4A
x4B
Exp 5
Exp 6
x5A
x6A
x5B
x6B
p1
Some distances: (MeV provides 11 metrics)
1. Euclidean: i6= 1 (xiA - xiB)2
2. Manhattan: i = 1 |xiA – xiB|
6
3. Pearson correlation
p0
Clustering Algorithms
Clustering Algorithms

Be weary - confounding computational
artifacts are associated with all clustering
algorithms. -You should always understand
the basic concepts behind an algorithm
before using it.

Anything will cluster! Garbage In means
Garbage Out.
Hierarchical Clustering
• IDEA: Iteratively combines genes into groups based on
similar patterns of observed expression
• By combining genes with genes OR genes with groups
algorithm produces a dendrogram of the hierarchy of
relationships.
• Display the data as a heatmap and dendrogram
• Cluster genes, samples or both
(HCL-1)
Hierarchical Clustering
Gene 1
Gene 2
Gene 3
Gene 4
Gene 5
Gene 6
Gene 7
Gene 8
Hierarchical Clustering
Gene 1
Gene 2
Gene 3
Gene 4
Gene 5
Gene 6
Gene 7
Gene 8
Hierarchical Clustering
Gene 1
Gene 2
Gene 4
Gene 5
Gene 3
Gene 8
Gene 6
Gene 7
Hierarchical Clustering
Gene 7
Gene 1
Gene 2
Gene 4
Gene 5
Gene 3
Gene 8
Gene 6
Hierarchical Clustering
Gene 7
Gene 1
Gene 2
Gene 4
Gene 5
Gene 3
Gene 8
Gene 6
Hierarchical Clustering
Gene 7
Gene 1
Gene 2
Gene 4
Gene 5
Gene 3
Gene 8
Gene 6
Hierarchical Clustering
Gene 7
Gene 1
Gene 2
Gene 4
Gene 5
Gene 3
Gene 8
Gene 6
Hierarchical Clustering
Gene 7
Gene 1
Gene 2
Gene 4
Gene 5
Gene 3
Gene 8
Gene 6
Hierarchical Clustering
H
L
Hierarchical Clustering
Genes
Samples
The Leaf Ordering Problem:
• Find ‘optimal’ layout of branches for a given dendrogram
architecture
• 2N-1 possible orderings of the branches
• For a small microarray dataset of 500 genes
there are 1.6*E150 branch configurations
Hierarchical Clustering
The Leaf Ordering Problem:
Hierarchical Clustering

Pros:
– Commonly used algorithm
– Simple and quick to calculate

Cons:
– Real genes probably do not have a
hierarchical organization
Self-Organizing Maps (SOMs)
A
Idea:
Place genes onto a
grid so that genes with
similar patterns of
expression are placed
on nearby squares.
B
C
D
a
c
bd
Self-Organizing Maps (SOMs)
A
IDEA:
Place genes onto a
grid so that genes with
similar patterns of
expression are placed
on nearby squares.
B
C
D
a
c
bd
Self-organizing Maps (SOMs)
Gene 1
Gene 2
Gene 3
Gene 4
Gene 5
Gene 6
Gene 7
Gene 8
Gene 9
Gene 10Gene 11
Gene 12
Gene 13
Gene 14
Gene 15
Gene 16
a_1hr
1
2
4
3
1
8
4
5
3
2
1
1
4
9
1
1
a_2hr
2
3
4
4
2
7
4
6
3
4
5
3
3
7
2
2
a_3hr
4
7
5
3
3
7
4
5
1
8
6
5
3
5
2
5
b_1hr
5
7
5
4
4
6
4
4
3
5
9
8
4
3
3
7
b_2hr
7
6
4
3
5
5
5
3
6
4
8
8
5
2
4
8
b_3hr
9
3
4
3
6
3
4
2
8
2
7
6
6
1
4
9
A
B
A
B
E
C
F
H
D
I
G
E
H
C
G
D
A
D
G
I
F
B
E
C
F
H
A
D
B
E
C
F
I
G
H
I
A
D
B
E
C
F
G
H
I
Self-organizing Maps (SOMS)
Gen es 11 and 12
Gen es 1, 1 6, and 5
Gen es 10
Gen e 1 5
A
D
G
B
E
H
C
F
I
Gen es 8
Gen es 6 and 14
Gen es 9 and 13
Gen es 3
Gen es 4, 7 an d 2
The Gene Expression Dynamics Inspector – GEDI
}
}
}
Samples
Group A
G en
en ee ss
G
Gene 1
Gene 2
Gene 3
Gene 4
Gene 5
Gene 6
…
Group C
Group B
1.5
1.4
1.7
1.2
.85
.65
.50
.55
2.5
2.8
2.7
2.1
.78
.95
.75
.45
1.1
1.2
1.0
1.3
.56
.62
.78
.89
.45
.23
.15
.05
.82
.71
.62
.49
.11
.16
.11
.95
2.2
4.5
6.7
6.2
2.2
2.5
2.8
2.9
.48
.90
1.5
1.8
2.1
2.0
1.9
1.6
4.2
4.8
5.2
5.5
2.5
2.6
2.0
1.9
1.2
1.1
1.6
2.9
1.1
1.8
1.9
1.4
1.7
1.2
1.1
1.6
GEDI’s Features:
•Allows for simultaneous analysis
or several time courses or datasets
•Displays the data in an intuitive
and comparable mathematically
driven visualization
•The same genes maps to the same
tiles
H
Group A
Group B
Group C
L
1
2
3
4
Software Demonstrations
MeV available at
http://www.tigr.org/software/tm4/mev.html
GEDI available at
http://www.chip.org/~ge/gedihome.htm
Comparison of GEDI vs. Hierarchical Clustering
Hierarchical clustering of random data
(GIGO)
G.E.D.I. allows the direct visual assessment of
the quality of conventional cluster analysis
From: CreateGEP_Journal.wpd, random_A
Questions