236632 – Topics in Microarray Data Analysis
Winter 2007-8
26.12.2007
Lecture #9:
Analysis of DNA Copy Numbers from Array-based Comparative Genomic
Hybridization (aCGH)
Lecturer: Zohar Yakhini
Scribes: Anat Eck & Lena Kleyner
Genomic instability is common in cancer. One form of the genomic instability is
changes in the copy number of particular regions such as deletions and amplifications.
Normal Karyotype
Colon Carcinoma HT-29
Figure 1
Comparing a normal karyotype to a carcinogenic cell line karyotype reveals differences in the copy
number and in the ploidy.
Some examples of changes in the DNA copy number in cancer:
 Retinoblastoma is a cancer of the retina. In the hereditary form of the disease,
all the cells of the body contain only one normal copy of the Rb gene. The
malignant tumor appears as a result of a deletion in 13q which leaves the cell
without a normal copy of Rb. The deletion happens in a high probability due
to the proximity of Rb to the telomere.
 A certain kind of breast cancer is caused by over expression and amplification
of the her2 (human epidermal growth factor receptor 2) gene. Herceptin is a
drug which blocks the her2 receptors but has severe side effects. A
CGH based test is used to determine whether the patient is her2 positive and
therefore will benefit from Herceptin.
Hence, measurement of DNA copy number plays an essential role in clinical
diagnostic and treatment. Nowadays, the modern medicine is on the edge of using
genetic information in order to choose the appropriate personal treatment.
CGH - Comparative Genomic Hybridization
CGH is a method which is used for detecting loss, gain and amplification of the copy
number at the levels of chromosomes throughout the whole genome.
There are two kinds of CGH:
1. Classic CGH method:
Two sets of samples are isolated and labeled - DNA from subject tissue
(usually cancer) is coloured in red and DNA from normal control tissue
(reference) is coloured in green. The mix is hybridized to normal
metaphase chromosomes which are used as a template.
The color ratio along the chromosomes is used to evaluate regions of DNA
gain or loss in the subject sample. A region coloured in red indicates of a
high amplification in the subject tissue (cancer), whereas green indicates
of a deletion in the subject tissue.
2. aCGH – array based CGH:
Similarly to the classic method, two sets of samples are isolated and
labeled, but the hybridization is performed using microarray.
In order to infer about the number of copies in the genome, an injective
binding is required.
The resolution of aCGH is higher than the one achieved in the classic
method, and is dependent on the probes uniqueness and the array size.
There are three different kinds of aCGH which differ in the content of the
array:
 BAC (bacterial artificial chromosome) - relative long fragments of
DNA which are cloned in bacteria. This technology is difficult to
maintain.
 cDNA- has better resolution, but is limited only for genes (since the
cDNA is produced in reverse transcription from mRNA).
 Oligonucleotide (60-mers) - the array size is larger (44 - 244K).
Enables to find deletion breaking points in the genome.
4
BAC array
[1998]
2
0
-2
0
1
2
3
4
5
6
7
8
7
x 10
cDNA array
[1999]
4
2
0
-2
0
1
2
3
4
5
6
7
8
7
x 10
oligo array
[2004]
4
2
0
-2
0
1
2
3
4
5
6
BT474 cell-line, chr 17 (x10Mbp)
7
8
7
x 10
Figure 2
In the above figure, the x axis describes the location on chromosome 17 and the y axis
describes the log  RED GREEN  value.
We can see the increasing resolution throughout the technique progress. The
amplification in chromosome 17 is also clearly seen.
The first aCGH utilized gene expression arrays, but since they were not
originally designed for that purpose the performance was relatively low.
The problems occurred when using gene expression arrays:
 Gene expression arrays are designated for RNA whereas CGH
measures DNA therefore the specificity is damaged.
 Gene expression arrays are limited to genes, whereas CGH
measures changes in the copy number throughout the whole
genome.

There are cases where the probe has no complementary sequence in
the genomic DNA. One example for that (the probe marked in red):
Exon1

Exon2
The specificity is damaged - a specific probe for cDNA may not be
specific anymore when we use the whole genomic DNA (e.g.
repetitive sequences).
Comparison between expression array and CGH specific array
The histograms below show a comparison between the performance of CGH using
gene expression array and the performance of CGH using a specific designed array
for detecting varying number of the X chromosome:
Theoretical
XY
XX
Log2
= -1.0
XX
XX
XXX
XX
XXXX
XX
XXXXX
XX
Log2
Log2
Log2
Log2
= 0.0
= 0.6
= 1.0
= 1.3

A
Human 1A
Expression
array
B




Slope=0.47





Number
of Probes
Measured Log2(Ratio)





















Measured Log2(Ratio)
C

D



Research
prototype
CGH array



Slope=0.96




Number
of Probes





Theoretical Log2(Ratio)


Measured Log2(Ratio) 

















Measured Log2(Ratio)
Barrett et al., PNAS 101:17765-17770, Dec 2004.







Theoretical Log2(Ratio)
Figure 3
The experiment included hybridizations of genomic DNA samples from a series of cell lines with
variable copy numbers of the X chromosome using 46,XX DNA as a reference. This was done in two
ways: using expression array (fig. A) and using a specific designed array (fig. C). Measured mean and
median fluorescence ratios of X-chromosome probes were plotted versus theoretical ratios for the
expression (fig. B) and CGH (fig. D) arrays.
A comparison of the median ratios for the X-chromosome probes from these hybridizations on the
expression array (373 probes) and the CGH array (4,878 probes) revealed that the slope increased from
0.47 to 0.96, which is much closer to the theoretical value – 1.
These data emphasize the usefulness of designing and selecting in situ synthesized oligonucleotide
probes for CGH assays.
Copy Number Polymorphism
Figure 4
A normal sample aCGH profile of chromosome 1 shows that beyond the expected noise, there are copy
number variations (the extreme points) which are common among the normal population.
CGH Data Analysis
Given a raw data matrix, our goal is to transform it into a corresponding step function.
This would enable the identification of genomic regions in which deletion or
amplification events have occured (aberration calling).
Figure 5
The figure above is an example for a corresponding step function to a given aCGH data.