Statistical Design and the Analysis of Gene Expression Microarray Experiments

1
Statistical Design and theAnalysis of Gene
Expression Microarray Experiments

• Gary A. Churchill
• The Jackson Laboratory
• Bar Harbor, ME 04609
• www.jax.org/research/churchill

2
Experimental Design forSpotted cDNA Microarrays
3
The design of an experiment is

• The set of treatments selected for comparison
• The specification of the units to which the
  treatments will be applied
• The rules by which the treatments are allocated
  to the experimental units
• The specifications of the measurements to be made

4
(No Transcript)
5
Loop Design
6
(No Transcript)
7

• Mice
• RNA extract
• Array

• E.g., strain
• Dye
• Labeled cDNA

8
The Dye-Swap Experiment
Sample1
Sample2
Array1
9
B
10
Repeated Dye-Flip
Sample1
Sample2
11
B
12
Replicated Dye-Swap
B1
A1
B2
A2
13

• Treatment
• Biological Replicate
• Technical Replicate
• Dye
• Array

14
Direct Comparison of Two Treatments
15
Direct Comparison of Multiple Samples
16
A Two Factor Design Sex x Strain
17
Indirect Comparisons via a Reference
Sample
18
Correlation in a Repeated Dye-Swap
array1 array2 array3 array4 R-gtG G-gtR R-gtG
G-gtR rep1 rep2 rep1 rep2 rep1 rep2 rep1 rep2 1.00
0.89 0.36 0.36 0.52 0.53 0.22 0.22 1.00
0.36 0.36 0.52 0.53 0.22 0.22
1.00 0.90 0.33 0.32 0.67 0.66
1.00 0.32 0.33 0.66 0.66 1.00 0.83 0.35 0.34
1.00 0.34 0.34 1.00 0.88 1.00
19
(No Transcript)
20
(No Transcript)
21
Sources of Variation

• Differences due to treatments
• Intrinsic biological variation (mouse)
• Technical variation in extraction and labeling of
  RNA samples
• Technical variation in hybridization
• Spot size variation
• Measurement error in scanning

22
Reducing Variation

• Replicated Spots
• Multiple Arrays per Sample
• Multiple Samples per Treatment Group
• Pooling

• Increased precision and quality control
• Estimate measurement error
• Estimate Biological Variation
• Reduce Biological Variation

23
Allocating Resourcesin a Microarray Experiment

• Measurement error
• Technical variance
• Biological variance
• effect of pooling
• Total variance

s2e s2A s2B
24
A Two Factor Design Sex x Diet with
2x replication
25
(No Transcript)
26
Individual Variation of Expression in Brain Cortex
I1
I2
I6
I3
I5
I4
27
(No Transcript)
28
Mouse Variation (CAMDA)
29
(No Transcript)
30
Summary Experimental Design

• Balance Dyes using Dye-Swap or Loops
• Use Independent Biological Replicates
• Use Technical Replicates
• Dont make one big pool