Microarray Data
Department of Statistics
University of Pune
Pune - 411007
Microarray Data
Microarray Data
The image of the microarray generated by the scanner is the
raw data of the experiment
Microarray Data
Microarray Data
The image of the microarray generated by the scanner is the
raw data of the experiment
For a single channel microarray, there is one image per array
while a two channel microarray produces two images per array,
one image per channel
Microarray Data
Microarray Data
The image of the microarray generated by the scanner is the
raw data of the experiment
For a single channel microarray, there is one image per array
while a two channel microarray produces two images per array,
one image per channel
Computer algorithms convert the image into the numerical
information that quantifies gene expression
Microarray Data
Microarray Data
The image of the microarray generated by the scanner is the
raw data of the experiment
For a single channel microarray, there is one image per array
while a two channel microarray produces two images per array,
one image per channel
Computer algorithms convert the image into the numerical
information that quantifies gene expression
First step of data analysis and has a major impact on the
quality of the data and the interpretation based on that
Microarray Data
Microarray Data
The image of the microarray generated by the scanner is the
raw data of the experiment
For a single channel microarray, there is one image per array
while a two channel microarray produces two images per array,
one image per channel
Computer algorithms convert the image into the numerical
information that quantifies gene expression
First step of data analysis and has a major impact on the
quality of the data and the interpretation based on that
Quantification of spotted cDNA or oligonucleotide array : The
initial step is to convert the digital TIFF(Tag Image File
Format) images of hybridization intensity of each channel on
each spot to numbers. This process involves following four
steps
Microarray Data
1 Identify the positions of the spots on the microarray
Microarray Data
1 Identify the positions of the spots on the microarray
2 For each spot, identify the pixels (picture elements) on the
image that are part of the spot
Microarray Data
1 Identify the positions of the spots on the microarray
2 For each spot, identify the pixels (picture elements) on the
image that are part of the spot
3 For each spot, identify nearby pixels that will be used for
background calculation
Microarray Data
1 Identify the positions of the spots on the microarray
2 For each spot, identify the pixels (picture elements) on the
image that are part of the spot
3 For each spot, identify nearby pixels that will be used for
background calculation
4 Calculate numerical information for the intensity of the spot,
the intensity of the background and quality control
information
Microarray Data
Microarray Data
1. Identify the positions of the spots on the microarray: Location
of each spot on the array is defined by assigning coordinates to the
center of each spot
Microarray Data
1. Identify the positions of the spots on the microarray: Location
of each spot on the array is defined by assigning coordinates to the
center of each spot
This process is known as gridding.
Microarray Data
1. Identify the positions of the spots on the microarray: Location
of each spot on the array is defined by assigning coordinates to the
center of each spot
This process is known as gridding.
Sometimes the grids come out naturally because there are
several pins on the cassette on the spotting robot; all the
spots in each grid being printed by the same pin.
Microarray Data
1. Identify the positions of the spots on the microarray: Location
of each spot on the array is defined by assigning coordinates to the
center of each spot
This process is known as gridding.
Sometimes the grids come out naturally because there are
several pins on the cassette on the spotting robot; all the
spots in each grid being printed by the same pin.
It is necessary to specify number of grids making up an array
and the parameters associated with the grids such as number
of grids in each direction, number of spots per grid in each
direction and the spacing between the grids.
Microarray Data
1. Identify the positions of the spots on the microarray: Location
of each spot on the array is defined by assigning coordinates to the
center of each spot
This process is known as gridding.
Sometimes the grids come out naturally because there are
several pins on the cassette on the spotting robot; all the
spots in each grid being printed by the same pin.
It is necessary to specify number of grids making up an array
and the parameters associated with the grids such as number
of grids in each direction, number of spots per grid in each
direction and the spacing between the grids.
The difficulties arising in identification involve, uneven grid
position, curve within a grid, uneven spot spacing, uneven
spot size.
Microarray Data
2. Segmentation: A process by which the software determines
which pixels (picture elements) in the area of a spot are part of the
spot and so their intensity will contribute to a quantitative
measurement of intensity at that spot. Commonly used methods
for segmentation
Microarray Data
2. Segmentation: A process by which the software determines
which pixels (picture elements) in the area of a spot are part of the
spot and so their intensity will contribute to a quantitative
measurement of intensity at that spot. Commonly used methods
for segmentation
Fixed circle segmentation: Fixed circle segmentation places a
circle of fixed size over the region of the spot and uses all the
pixels in the circle as part of the spot. The problem with fixed
circle segmentation is that it gives inaccurate results if the
spots are of different sizes which is usually the case on most
microarrays
Microarray Data
2. Segmentation: A process by which the software determines
which pixels (picture elements) in the area of a spot are part of the
spot and so their intensity will contribute to a quantitative
measurement of intensity at that spot. Commonly used methods
for segmentation
Fixed circle segmentation: Fixed circle segmentation places a
circle of fixed size over the region of the spot and uses all the
pixels in the circle as part of the spot. The problem with fixed
circle segmentation is that it gives inaccurate results if the
spots are of different sizes which is usually the case on most
microarrays
Variable circle segmentation: Variable circle segmentation fits
a circle of variable size onto the region containing the spot.
This method is able to resolve spots of different sizes, but
performs less well on irregularly shaped spots
Microarray Data
Histogram segmentation: Histogram segmentation fits a circle
over the region of the spot as well as background and then
looks at a histogram of the intensities of the pixels in the
spot. The brightest and dimmest pixels are not used in the
quantification of spot intensity. Histogram segmentation
produces reliable results for irregularly shaped spots
Microarray Data
Histogram segmentation: Histogram segmentation fits a circle
over the region of the spot as well as background and then
looks at a histogram of the intensities of the pixels in the
spot. The brightest and dimmest pixels are not used in the
quantification of spot intensity. Histogram segmentation
produces reliable results for irregularly shaped spots
Adaptive shape segmentation: Algorithm requires a smaller
number of seed pixels in the centre of each spot to start. It
then extends the regions of each spot by adjoining pixels that
are similar in intensity to their neighbors
Microarray Data
3. Estimation of background intensity : Although the scanner is
supposed to pick up light emitted by the target cDNAs bound to
their complementary spots, it will inevitably also pick up light from
various other sources, including the labeled sample hybridizing non
specifically to the glass slide, that is unwashed labeled sample
adhering to the slide, various chemicals used in processing the slide
and even the slide itself
This extra light is called the background intensity
Microarray Data
3. Estimation of background intensity : Although the scanner is
supposed to pick up light emitted by the target cDNAs bound to
their complementary spots, it will inevitably also pick up light from
various other sources, including the labeled sample hybridizing non
specifically to the glass slide, that is unwashed labeled sample
adhering to the slide, various chemicals used in processing the slide
and even the slide itself
This extra light is called the background intensity
The background intensities are expected to be low, however,
the main concern is that the spot intensities also contain a
certain amount of this background fluorescence
Microarray Data
3. Estimation of background intensity : Although the scanner is
supposed to pick up light emitted by the target cDNAs bound to
their complementary spots, it will inevitably also pick up light from
various other sources, including the labeled sample hybridizing non
specifically to the glass slide, that is unwashed labeled sample
adhering to the slide, various chemicals used in processing the slide
and even the slide itself
This extra light is called the background intensity
The background intensities are expected to be low, however,
the main concern is that the spot intensities also contain a
certain amount of this background fluorescence
It may be substantial in some spots and may vary with the
location of the spot
Microarray Data
It is customary therefore to estimate background intensity
from data. Assuming that the spot signal intensity is an
additive combination of the true spot intensity and the
background, the background intensity is subtracted from the
raw spot intensity values to yield a set of background-adjusted
spot intensity values
Microarray Data
It is customary therefore to estimate background intensity
from data. Assuming that the spot signal intensity is an
additive combination of the true spot intensity and the
background, the background intensity is subtracted from the
raw spot intensity values to yield a set of background-adjusted
spot intensity values
Global background adjustment: A very simple estimate of the
background is the average intensity of all the pixels not
belonging to spots
Microarray Data
It is customary therefore to estimate background intensity
from data. Assuming that the spot signal intensity is an
additive combination of the true spot intensity and the
background, the background intensity is subtracted from the
raw spot intensity values to yield a set of background-adjusted
spot intensity values
Global background adjustment: A very simple estimate of the
background is the average intensity of all the pixels not
belonging to spots
This approach is rarely effective, because the background is
often not uniform over the entire microarray
Microarray Data
It is customary therefore to estimate background intensity
from data. Assuming that the spot signal intensity is an
additive combination of the true spot intensity and the
background, the background intensity is subtracted from the
raw spot intensity values to yield a set of background-adjusted
spot intensity values
Global background adjustment: A very simple estimate of the
background is the average intensity of all the pixels not
belonging to spots
This approach is rarely effective, because the background is
often not uniform over the entire microarray
Spot background adjustment: Estimated as the average
intensity of all the pixels in the vicinity of the spot
Spot background is subtracted from the spot intensity value
to yield a spot background-adjusted spot intensity value
Microarray Data
In principle, spot intensity should be larger than background
intensity
Microarray Data
In principle, spot intensity should be larger than background
intensity
However, due to some problems, on some spots background
intensity can exceed spot intensity leading to a negative value
for background-adjusted spot intensity
Microarray Data
In principle, spot intensity should be larger than background
intensity
However, due to some problems, on some spots background
intensity can exceed spot intensity leading to a negative value
for background-adjusted spot intensity
Such spots are flagged and usually omitted from further
analysis
Microarray Data
In principle, spot intensity should be larger than background
intensity
However, due to some problems, on some spots background
intensity can exceed spot intensity leading to a negative value
for background-adjusted spot intensity
Such spots are flagged and usually omitted from further
analysis
As this is not desirable, sometimes a small additional
adjustment is made if T is a low percentile of the spot
intensity values (for example, the fifth percentile), the
background and threshold adjusted spot intensity value, b is
defined as
b = max (Background-adjusted spot intensity, T).
Microarray Data
4) Calculation of numerical information at the spot (i) Signal
mean: mean of the pixel intensities comprising the spot
Microarray Data
4) Calculation of numerical information at the spot (i) Signal
mean: mean of the pixel intensities comprising the spot
(ii) Background mean: mean of the pixel intensities
comprising the background of the spot.
Microarray Data
4) Calculation of numerical information at the spot (i) Signal
mean: mean of the pixel intensities comprising the spot
(ii) Background mean: mean of the pixel intensities
comprising the background of the spot.
(iii) Signal median: median of the pixel intensities comprising
the spot
Microarray Data
4) Calculation of numerical information at the spot (i) Signal
mean: mean of the pixel intensities comprising the spot
(ii) Background mean: mean of the pixel intensities
comprising the background of the spot.
(iii) Signal median: median of the pixel intensities comprising
the spot
(iv) Background median: median of the pixels comprising the
background
Microarray Data
4) Calculation of numerical information at the spot (i) Signal
mean: mean of the pixel intensities comprising the spot
(ii) Background mean: mean of the pixel intensities
comprising the background of the spot.
(iii) Signal median: median of the pixel intensities comprising
the spot
(iv) Background median: median of the pixels comprising the
background
(v) Signal standard deviation: standard deviation of the pixel
intensities comprising the spot
Microarray Data
4) Calculation of numerical information at the spot (i) Signal
mean: mean of the pixel intensities comprising the spot
(ii) Background mean: mean of the pixel intensities
comprising the background of the spot.
(iii) Signal median: median of the pixel intensities comprising
the spot
(iv) Background median: median of the pixels comprising the
background
(v) Signal standard deviation: standard deviation of the pixel
intensities comprising the spot
(vi) Background standard deviation: standard deviation of the
pixel intensities comprising the background.
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
(viii) Number of pixels: number of pixels comprising the spot
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
(viii) Number of pixels: number of pixels comprising the spot
(ix) Flag: a variable that is 0 if the spot is good, and will take
different values if the spot is not good
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
(viii) Number of pixels: number of pixels comprising the spot
(ix) Flag: a variable that is 0 if the spot is good, and will take
different values if the spot is not good
Different image-processing software use different flag values
for different problems
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
(viii) Number of pixels: number of pixels comprising the spot
(ix) Flag: a variable that is 0 if the spot is good, and will take
different values if the spot is not good
Different image-processing software use different flag values
for different problems
Typical problems are
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
(viii) Number of pixels: number of pixels comprising the spot
(ix) Flag: a variable that is 0 if the spot is good, and will take
different values if the spot is not good
Different image-processing software use different flag values
for different problems
Typical problems are
Bad spot: pixel standard deviation is very high relative to the
pixel mean
Microarray Data
(vii) Diameter: number of pixels across the width of the spot
(viii) Number of pixels: number of pixels comprising the spot
(ix) Flag: a variable that is 0 if the spot is good, and will take
different values if the spot is not good
Different image-processing software use different flag values
for different problems
Typical problems are
Bad spot: pixel standard deviation is very high relative to the
pixel mean
Negative spot: signal of the spot is less than the signal of the
background
Microarray Data
Microarray Data
Microarray Data