The Genome Access Course Microarray Informatics

1
TheGenomeAccessCourseMicroarrayInformatics
Maple Leaves
2
Yeast Microarray
3
From Brown PO, Botstein D. Exploring the new
world of the genome with DNA microarrays. Nat
Genet. 1999 21 33-7
4
From Brown PO, Botstein D. Exploring the new
world of the genome with DNA microarrays. Nat
Genet. 1999 21 33-7
5
From Brown PO, Botstein D. Exploring the new
world of the genome with DNA microarrays. Nat
Genet. 1999 21 33-7
6
(No Transcript)
7
(No Transcript)
8
Two-Color Detection

• Cy3 and Cy5
• Forgiving of array imperfections
• Generates robust ratio data
• Scalable

9
Systems

• Affymetrix
• Axon
• Home-grown

10
Terminology

• Feature array element
• Probe a feature corresponding to a defined
  sequence (synthetic oligos or cDNAs)
• Target nucleic acid pool of unknown sequence

11
Issues

• Array Fabrication
• Probe Preparation
• Hybridization
• Image Analysis
• Data Visualization/Analysis
• Data Storage

12
Image Analysis

• Feature Identification
• Background
• Median vs. mean
• GenePix

13
Data Analysis

• Normalization
• Replicates
• Expression Analysis
• Clustering
• Hierarchical Clustering
• k-Means
• Self-Organizing Maps

14
Data Storage
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
Normalization

• Correct for systematic bias in data
• Attempts to remove non-biological influences from
  biological data
• Provides a baseline for comparison between
  microarrays
• Possibly compare data from one platform to another

23
Sources of Variation

• Printing and/or tip problems
• Labeling and dye effects (differing amounts of
  RNA labeled between the 2 channels)
• Differences in the power of the two lasers (or
  other scanner problems)
• Difference in DNA concentration on arrays (plate
  effects)
• Spatial biases in ratios across the surface of
  the microarray due to uneven hybridization

24
An Example

• In array 1, the Cy5 dye labels twice as
  efficiently of the is 2 times better than for the
  Cy3 dye.
• Everything else is the same
• Ratios will be 2 instead of 1
• In array 2, the Cy5 dye labels only 1.5 times as
  efficiently
• Ratios will be 1.5 instead of 1

25
Methods

• Housekeeping genes
• Spiked-in controls
• Mean/median log ratio adjustment
• Loess curve fit
• Local versions of above two
• Use of reverse-labeled replicates (dye swap
  experiments)

26
Methods Normalization Factor/Function Calculation

• Global mean or median normalization This method
  calculates a normalization factor based on the
  selected elements, as either the mean or median
  log ratio of those selected elements. This value
  is then subtracted from the log ratio of the
  elements to which this normalization factor is to
  be applied. If you plot a histogram of log
  ratios, you get a roughly normal distribution.
  What this method does is simply shift that
  distribution along the x-axis, so that it is
  centered around zero.
• Intensity-dependent normalization A function is
  generated, using the selected elements, that is
  intensity-dependent. This is usually done as a
  loess fit to a plot of log ratio vs log (mean
  intensity). This function is then applied to the
  data.

27
Background Subtraction

• Element Intensity - Element Background Intensity
  Background Corrected Intensity
• Can be local or global
• Sensitive to spot morphologies

28
Global Mean Normalization

• Element Intensity / Global Mean Element Intensity
  Global Mean Normalized Intensity
• Performed for both sets of intensities separately

29
Local Mean Normalization across Microarray Surface

• Corrects spatial artifacts
• Requires x y coordinates
• Element Intensity/ Local Mean Intensity Local
  Mean Normalized Intensity

30
Logarithmic Transformation

• Perform a logarithmic transformation of all
  intensities
• Base 10 is common

31
Calculate Mean Log(Intensities) and Log(Ratios)

• X axis is the mean gene expression level in the
  two samples
• Mean (Log(Intensity)) Geomentric Mean Intensity
• Y axis is a measure of differential gene
  expression between the two samples
• Log(ONE/TWO) Log(ONE) - Log(TWO)

32
Local Mean Normalization across Element Signal
Intensity

• Generates a loess fit from local mean intensity
  levels
• Log (Ratio) - Local Mean Intensity Residual
  Corrected Log (Ratio)

33
Local Variance Correction across Element Signal
Intensity

• Loess fit from local standard deviations
  Corrected Log (Ratio) / Local Stadard Deviation
  Local Z-Score