My Masters Thesis Defense in Computational Science at San Diego State University (SDSU). Thesis title: "Microarray Analysis of the Effects of Rosiglitazone on Gene Expression in Neonatal Rat Ventricular Myocytes", Fall 2009.
Privatization and Disinvestment - Meaning, Objectives, Advantages and Disadva...
Effects of Rosiglitazone on Gene Expression in Neonatal Rat Heart Cells
1. Microarray Analysis of the Effects of Rosiglitazone
on Gene Expression in Neonatal Rat Ventricular
Myocytes
Elliot Kleiman
San Diego State University
Masters Thesis Defense in Computational Science
September 17, 2009
7. Diabetes
What is it?
How many people are affected?
Cardiovascular complications
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 3 / 41
8. Diabetes
What is it?
How many people are affected?
Cardiovascular complications
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 3 / 41
9. Diabetes
What is it?
How many people are affected?
Cardiovascular complications
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 3 / 41
10. Rosiglitazone
Prescription drug which lowers blood sugar levels
Avandia®(1999, GlaxoSmithKline), U.S. patent 2012
Controversial drug
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 4 / 41
11. Rosiglitazone
Prescription drug which lowers blood sugar levels
Avandia®(1999, GlaxoSmithKline), U.S. patent 2012
Controversial drug
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 4 / 41
12. Rosiglitazone
Prescription drug which lowers blood sugar levels
Avandia®(1999, GlaxoSmithKline), U.S. patent 2012
Controversial drug
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 4 / 41
13. Previous work
Shah et al. 2002 (M.Sci Biology, SDSU) found that Rosiglitazone:
Improves cardiac contractility by enhancing cytosolic calcium
removal
Increases SERCA2 mRNA, protein, and promoter activity
Increases NFκB promoter and IL-6 protein secretion
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 5 / 41
14. Previous work
Shah et al. 2002 (M.Sci Biology, SDSU) found that Rosiglitazone:
Improves cardiac contractility by enhancing cytosolic calcium
removal
Increases SERCA2 mRNA, protein, and promoter activity
Increases NFκB promoter and IL-6 protein secretion
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 5 / 41
15. Previous work
Shah et al. 2002 (M.Sci Biology, SDSU) found that Rosiglitazone:
Improves cardiac contractility by enhancing cytosolic calcium
removal
Increases SERCA2 mRNA, protein, and promoter activity
Increases NFκB promoter and IL-6 protein secretion
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 5 / 41
16. Previous work
Shah et al. 2002 (M.Sci Biology, SDSU) found that Rosiglitazone:
Improves cardiac contractility by enhancing cytosolic calcium
removal
Increases SERCA2 mRNA, protein, and promoter activity
Increases NFκB promoter and IL-6 protein secretion
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 5 / 41
17. Current thesis work
Are there other genes affected by rosiglitazone in addition to
SERCA2?
Can we:
identify these genes?
determine their functional relationships?
classify these genes as early or late responders over time?
How to implement these objectives?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 6 / 41
18. Current thesis work
Are there other genes affected by rosiglitazone in addition to
SERCA2?
Can we:
identify these genes?
determine their functional relationships?
classify these genes as early or late responders over time?
How to implement these objectives?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 6 / 41
19. Current thesis work
Are there other genes affected by rosiglitazone in addition to
SERCA2?
Can we:
identify these genes?
determine their functional relationships?
classify these genes as early or late responders over time?
How to implement these objectives?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 6 / 41
20. Current thesis work
Are there other genes affected by rosiglitazone in addition to
SERCA2?
Can we:
identify these genes?
determine their functional relationships?
classify these genes as early or late responders over time?
How to implement these objectives?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 6 / 41
21. Current thesis work
Are there other genes affected by rosiglitazone in addition to
SERCA2?
Can we:
identify these genes?
determine their functional relationships?
classify these genes as early or late responders over time?
How to implement these objectives?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 6 / 41
22. Current thesis work
Are there other genes affected by rosiglitazone in addition to
SERCA2?
Can we:
identify these genes?
determine their functional relationships?
classify these genes as early or late responders over time?
How to implement these objectives?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 6 / 41
23. Gene expression primer
Replication
Genes
DNA
Transcription
(RNA synthesis)
Gene
Expression RNA
Translation
(Protein synthesis)
Phenotype PROTEIN
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 7 / 41
24. Experimental approach
DNA microarrays, useful why?
because one can measure the gene expression levels of thousands
of genes simultaneously
because measuring the levels of mRNA is easier than measuring
levels of proteins
because mRNA is a good surrogate marker for protein (or is it?)
because when you don’t have a hypothesis, microarrays can help
you find one
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 8 / 41
25. Experimental approach
DNA microarrays, useful why?
because one can measure the gene expression levels of thousands
of genes simultaneously
because measuring the levels of mRNA is easier than measuring
levels of proteins
because mRNA is a good surrogate marker for protein (or is it?)
because when you don’t have a hypothesis, microarrays can help
you find one
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 8 / 41
26. Experimental approach
DNA microarrays, useful why?
because one can measure the gene expression levels of thousands
of genes simultaneously
because measuring the levels of mRNA is easier than measuring
levels of proteins
because mRNA is a good surrogate marker for protein (or is it?)
because when you don’t have a hypothesis, microarrays can help
you find one
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 8 / 41
27. Experimental approach
DNA microarrays, useful why?
because one can measure the gene expression levels of thousands
of genes simultaneously
because measuring the levels of mRNA is easier than measuring
levels of proteins
because mRNA is a good surrogate marker for protein (or is it?)
because when you don’t have a hypothesis, microarrays can help
you find one
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 8 / 41
28. Experimental approach
DNA microarrays, useful why?
because one can measure the gene expression levels of thousands
of genes simultaneously
because measuring the levels of mRNA is easier than measuring
levels of proteins
because mRNA is a good surrogate marker for protein (or is it?)
because when you don’t have a hypothesis, microarrays can help
you find one
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 8 / 41
30. Bead design
BEAD DESIGN
Labelled
cRNA
Address Probe
29b 50b
Gene-speci c probes are concatenated
with a short "address sequence."
Source: Illumina.com
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 10 / 41
31. Materials & Methods
Drug = rosiglitazone
Control = dimethylsulfoxide (DMSO)
Two samples of ≈100 newborn (neonatal) rats
isolated and cultured neonatal rat ventricular myocytes (NRVMs)
48 arrays or 4 Illumina RatRef-12 Expression BeadChips
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 11 / 41
32. Study Design
Table: 12×2 Factorial Design
Timea
(hour)
0b ½ 1 2 4 6 8 12 18 24 36 48
DMSO -c +d + + + + + + + + + +
Drug
Rosiglitazone - + + + + + + + + + + +
DMSO, dimethylsulfoxide.
a
Exposure time to drug treatment.
b
Untreated RNA.
c
No drug administered.
d
Drug administered.
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 12 / 41
33. Array hybridization layout
Sample 1 Sample 2
06/21/07 06/21/07 07/10/07 07/10/07
A R 0.5hr A U A R 0.5hr A U
D 48hr B U B D 48hr B U B
C D 36hr C R 48hr C D 36hr C R 48hr
D 24hr D R 36hr D D 24hr D R 36hr D
E D 18hr E R 24hr E D 18hr E R 24hr
D12 hr F R 18hr F D12 hr F R 18hr F
G D 8hr G R 12hr G D 8hr G R 12hr
D 6hr H R 8hr H D 6hr H R 8hr H
I D 4hr I R 6hr I D 4hr I R 6hr
D 2hr J R 4hr J D 2hr J R 4hr J
K D 1hr K R 2hr K D 1hr K R 2hr
D 0.5hr L R 1hr L D 0.5hr L R 1hr L
1677718214 1677718210 1677718217 1677718209
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 13 / 41
35. Data Analysis
Data analysis goal: to find an association between treatment
condition and gene expression
Common gene selection strategies:
Fold change
Parametric test: two sample t-test
Non-parametric tests: rank sum, signed-rank tests
ANOVA
Permutation or bootstrap resampling
. . . zillions of others!
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 15 / 41
36. Data Analysis
Data analysis goal: to find an association between treatment
condition and gene expression
Common gene selection strategies:
Fold change
Parametric test: two sample t-test
Non-parametric tests: rank sum, signed-rank tests
ANOVA
Permutation or bootstrap resampling
. . . zillions of others!
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 15 / 41
37. Data Analysis
Data analysis goal: to find an association between treatment
condition and gene expression
Common gene selection strategies:
Fold change
Parametric test: two sample t-test
Non-parametric tests: rank sum, signed-rank tests
ANOVA
Permutation or bootstrap resampling
. . . zillions of others!
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 15 / 41
38. Data Analysis
Data analysis goal: to find an association between treatment
condition and gene expression
Common gene selection strategies:
Fold change
Parametric test: two sample t-test
Non-parametric tests: rank sum, signed-rank tests
ANOVA
Permutation or bootstrap resampling
. . . zillions of others!
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 15 / 41
39. Linear models of microarrays (LIMMA)
Linear Model
log(ygi ) = µg + βgR xRi + βgD xDi + βgR:D xRi xDi + gi (1)
Idea: use a linear model to parameterize the effects of drug and
time from our factorial designed experiment
Source: Smyth, Limma (2004)
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 16 / 41
40. Moderated, bayesian t-test
Moderated t-statistic
2 2
d0 s0 − dg sg
2
sg =
d0 + dg
(2)
∗ βg
tg =
sg ug
2 2
Std.Err used in test-statistic is a weighted average of s0 + sg
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 17 / 41
41. Significant contrasts of interest
Table: Numbers of genes regulated during significant exposure times to
rosiglitazone vs. DMSO in NRVMs
Significant exposure times for rosiglitazone vs. DMSO
(hour)
2 4 6 8 12 18 24 36 48
a
-1 0 0 0 0 0 0 2 8 9
No. genes
regulated 0b 22516 22513 22514 22513 22506 22506 22498 22491 22491
1c 1 4 3 4 11 11 17 18 17
a
Numbers of genes down-regulated.
b
Numbers of genes unchanged.
c
Numbers of genes up-regulated.
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 18 / 41
47. What is a biological pathway?
Biological process: The set of all molecules required to perform a
biological function
Biological pathway: The set of all molecular interactions that belong to
a biological process
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 24 / 41
48. Overrepresented KEGG pathways
PPAR signaling
Fatty acid metabolism
Synthesis and degradation of ketone bodies
Valine, leucine, and isoleucine degradation
Butanoate metabolism
Bile acid metabolism
ATP binding cassette transporters, general
biol. pathway theme: fatty acid and lipid metabolism and mitochondrial energy transfer
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 25 / 41
51. Gene ontology, what is it?
structured vocabulary for describing genes and gene products
molecular function (what it does)
biological process (how it contributes)
cellular component (where it does it)
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 28 / 41
52. Hypergeometric testing
test of association between two categories of interest (equivalent
to Fisher’s Exact test)
used to assess the over-representation of GO terms
how many genes in the universe(array) are annotated at a given
term?
how many of those are also in the set of interesting genes?
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 29 / 41
53. Hypergeometric testing case example
Universe = 1000 genes, 400 are DE, GO term has 40 annotations
What is the Prob that 10 of the 40 genes in GO term are also in the set
of DE?
DE DE Total
In GO term 10 30 40
On Array 390 570 960
Total 400 600 1000
Falcon, GOstats, 2007
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 30 / 41
54. Hypergeometric random variable
e.g., sampling balls from an urn model without replacement each trial
is dependent on the previous one
Hypergeometric random variable
k N−k
y n−y
P(y) = N
n
where N = population size k = number of population successes n =
sample size y = number of sample successes
from prev slide we would have,
400 600
10 30
P(10) = 1000
= 0.99
40
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 31 / 41
55. Overrepresented GO terms
Biological process Cellular component Molecular function
primary metabolic lipid particle catalytic activity
process mitochondrion electron carrier activity
lipid metabolic process mitochondral membrane transferase activity
cellular lipid metabolic mitochondral inner transferring acyl groups
process membrane acyltransferase activity
oxidation reduction nuclear oxidoreductase activity
response to drug envelope-endoplasmic
reticulum network
gene ontology theme: energetic and metabolism activities
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 32 / 41
57. Angptl4 & Adfp
Angptl4, Angiopoietin-like protein 4 Adfp, Adipose differentiation protein
up-regulated 3 to 7 fold up-regulated 1.5 to 1.7 fold
potent inhibitor of LPL plays a key role in formation of lipid
plays key role in modulating cardiac droplets
substrate metabolism lipid droplet associated protein
decreases TG delivery to heart for FA β adipocyte differentiation
oxidation responsible for increase in subcutaneous
tissue mass observed in rosiglitazone
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 34 / 41
58. Lipid & energy metabolism in cardiomyocytes
Lipoprotein secretion
VLDL Albumin
Chylomicrons LDL
TG FFA
TG TG
FATP CD36
FFA
sis
LACS TG synthe TG
Intermembrane
Mitochondria Fatty Acyl-CoA space
lipolysis
Outer membrane CPT-I
Carnitine Acylcarnitine
Inner membrane
CPT-II CACT
FADH
FAD +
Acyl-CoA Coenzyme A
Enoyl-CoA UCP2,
LCAD
UCP3
Enoyl-CoA H 2O Energy uncoupling
hydratase
3-OH-acyl-CoA
NAD +
HAD
NAD+H + CO 2 ATP
TCA cycle
n
3-ketoacyl-CoA
ai
CoA-SH
ch
ry
to
Thiolase
ira
Acyl-CoA + Acetyl-CoA
sp
re
n
t ro
ec
El
Yang & Li 2007
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 35 / 41
59. Actions of PPARγ in FA trapping
Semple, 2006
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 36 / 41
60. Molecular mechanisms of TZDs
PPARg
Coactivator binding site
Coactivator fragment Ligand binding site
Transactivation Transrepression
TZD TZD
Ligands Ligands
PPARg PPARg
Ligand activation Ligand activation
Coactivator
PPARg
Cofactor recruitment
p65 p50 Fos Jun STAT1 STAT3
PPARg RXR X X X
PPAR target genes
PPRE PPRE
NF- kB-RE TRE ISGF-RE
Hannele Yki-Jarvinen, (SDSU)
Elliot Kleiman 2004 Microarray Analysis Sept. 17, 2009 37 / 41
61. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
62. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
63. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
64. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
65. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
66. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
67. Limitations
Drug
No PCR validation
No assay for PPARγ levels
technician-level variation
Limitations of the array technology
Sample size of 2
No db/db disease model
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 38 / 41
68. Acknowledgements
SDSU UC San Diego Northwestern UC Riverside
Paul Paolini Gary Hardiman University Thomas Girke
Jose Castillo Roman Sasik Denise Scholtens
Peter Salamon Charles Berry Pan Du
James Otto Jennifer Lapira Simon Lin
Frank Gonzales
Lynelle Garnica
Kirubel
Gebresenbet
Magda Nemeth
David Torres
Evri Linux administration Illumina, Inc. EMD Biosciences
Seth Falcon Greg Chandler Andrew Carmen Huda Shubeita
Elliot Kleiman (SDSU) Microarray Analysis Sept. 17, 2009 39 / 41