cDNA Microarray Analysis of BCell Chronic Lymphocytic Leukemia

1
cDNA Microarray Analysis of B-Cell Chronic
Lymphocytic Leukemia

• Lynne V. Abruzzo, M.D., Ph.D.
• Department of Hematopathology
• Kevin R. Coombes, Ph.D.
• Department of Biostatistics

2
Outline

• Introduction
• Image Quantification
• Data Preprocessing
• Sources of Variability
• Gene Expression Profiling of B-Cell Chronic
  Lymphocytic Leukemia
• Differentially Expressed Genes
• Annotation and Interpretation

3
B-Cell Chronic Lymphocytic Leukemia(B-CLL)

• Epidemiology
• most common leukemia in the Western Hemisphere
• median age 65 years
• male predominance (MF 21)
• Morphology
• small, mature-appearing lymphoid cells
• bone marrow, peripheral blood, lymph nodes
• Immunophenotype
• pan B-cell antigens
• CD5
• surface IgM IgD

4
Genetic Findings in B-CLL

• Immunoglobulin Genes
• clonal Ig heavy and light chain rearrangements
• somatic mutation of VH genes (60)
• unmutated VH genes (40)
• Cytogenetic Abnormalities
• deletions at 13q14 (50)
• trisomy 12 (20)
• 11q22-23 (20)
• deletions at 17p13 (10)
• deletions at 6q21 (5)

5
Applications of Microarrays to Tumor
Classification

• Class Prediction
• supervised assign samples to known classes
• ex. support vector machines, discriminant
  analysis,
• k-nearest neighbor, etc.
• Class Discovery
• unsupervised find new classes from the data
• ex. hierarchical clustering, principal component
  analysis,
• self organizing maps, etc.
• Gene Discovery
• identify differentially expressed genes

6
Membrane cDNA Arrays

• Research Genetics GeneFilter arrays
• versions I-VI (GF200-GF205)
• 5,184 ESTs per array
• I.M.A.G.E./LLNL cDNA clones
• 0.5 ng cDNA per spot

7
Overview of the Method
Cells
RNA Isolation
33P dCTP
Reverse Transcription
Imaging with PhosphorImager
Hybridization
Membrane
8
Image Quantification

• Alignment
• Quantification
• Intensity
• Background
• Area
• Variance of pixel intensity

9
Alignment
10
Quantified Data
thousands of rows...
11
Data Preprocessing

• Goal
• Process raw quantification data so they can be
  compared across microarray experiments
• Preprocessing Steps
• Background correction
• Normalization
• Thresholding
• Log transformation

12
Background Correction

• Mean, median, or something else?
• Local, global, or something in between?
• Where is it measured?
• Ring around the spot
• Corners between spots
• Edge of the array

13
Plot of Signal vs. Background
14
Normalization

• To what level?
• Housekeeping genes
• Spiked controls
• Mean or median of all spots
• Mean or median of expressed spots
• On what scale?
• Entire array
• Patch-by-patch (pin effects)

15
Thresholding

• How small a signal can we detect?
• No absolute answer
• Relative answers
• Signal to noise ratio at least 2 (or 3, or )
• Greater than 95 (or 99, or ) of known blank
  spots or negative controls
• To estimate ratios conservatively, replace small
  values by an estimate of the detection threshold.

16
Variability of Expression Increases with Mean
Intensity
17
Variability of Expression Changes with Mean Log
Intensity
18
Sources of Variability

• Biologic variability
• Technical variability
• array fabrication
• identity of clones, mechanical printing errors,
  etc.
• target preparation
• sample homogeneity, RNA quality, etc.
• target hybridization
• labeling efficiency, membrane reuse, etc.
• imaging, image analysis, image quantification
• image alignment, membrane distortion, etc.

19
Experiments to Assess Variability

• Design
• 15 hybridizations performed with 2 extractions of
  total RNA from the GA-10 cell line (Burkitt
  lymphoma)
• total RNA reverse transcribed and labeled with
  33P
• 4 Research Genetics GF200 GeneFilters membrane
  arrays
• Each membrane reused up to 6 times
• PhosphorImager screens exposed from 2 - 26 hours
• Images independently quantified twice
• Data
• Total data set 44 quantifications of 22 images
• Images quantified, background corrected,
  normalized, thresholded, and log transformed

20
Contributions to Variance
21
Percent Contributions to Variance
22
Hierarchical Clustering GA-10 vs. NBC
23
Principal Components AnalysisGA-10 vs. NBC
24
Gene Expression Profiling of B-Cell Chronic
Lymphocytic Leukemia
25
Experimental Design

• Samples
• CD19-positive peripheral blood lymphocytes
• 6 untreated patients with B-CLL
• 6 normal donors
• Hybridizations
• 33P-labeled cDNA prepared from 3 ?g total RNA
• GeneFilters versions I-VI (31,104 clones)
• Data
• total data set 72 quantified images

26
Hierarchical ClusteringB-CLL vs. NBC
27
Principal Components AnalysisB-CLL vs. NBC
28
Scatter Plot of Mean Log IntensityB-CLL vs. NBC
29
(No Transcript)
30

Rotated Scatter PlotB-CLL vs. NBC
31

Rotated Scatter PlotB-CLL vs. NBC
32

Rotated Scatter PlotB-CLL vs. NBC
33
Number of Genes Identified as Differentially
Expressed
Single test p 0.005 expect about 25 genes by
chance
34
Hierarchical Clustering
35
Principal Components Analysis
36
Differentially Expressed GenesB-CLL vs. NBC
and hundreds more...
37
Differentially Expressed GenesB-CLL vs. NBC
38
Annotation

• Manufacturer usually supplies a file identifying
  the spots on the microarray
• IMAGE clone ID
• GenBank accession number
• From this, can find
• UniGene cluster ID
• Gene name and symbol
• Chromosome location
• Gene function

39
Functional Categories of Differentially Expressed
Genes
40
Gene ClustersB-CLL vs. NBC
41
Collaborators

• UT M.D. Anderson Cancer Center
• Hematopathology
• Lynn Barron
• Biostatistics
• Sang-joon Lee
• Leukemia
• Michael Keating

• University of Maryland
• Pathology
• W. Edward Highsmith
• Tammy Krogmann

42
Web Site

• http//www.mdanderson.org/
• depts/cancergenomics/reports.html

43
(No Transcript)