Differential Gene Expression: Ischemic vs. Nonischemic

1
Differential Gene Expression Ischemic vs.
Nonischemic

• Jing Hu
• Dongmei Li
• Shuyan Wan
• Richard Yamada
• Jeong-Me Yoon
• Zailong Wang (Mentor)

2
Outline for Our Talk

• Introduction and summary of previous work
  (Richard)
• Exploratory Analysis of Data (Jeong-Mi)
• Statistical Methods (Shuyan)
• Selected Gene Analysis (Jing)
• Conclusions and Further Work (Dongmei)

3
Human Heart Function
4
Arteries
5
What is Ischemic Cardiomyopathy?

• Ischemic Lack of Blood and Oxygen
• Cardio Refers to the Heart
• Myopathy Muscle Related Disease
• ischemic cardiomyopathy is a medical term that
  doctors use to describe patients who have
  congestive heart failure that is a result of
  coronary artery disease. (coronary arteries are
  blocked)

6
Ischemic Cardiac Myopathy

• Risk Factors genetics, smoking, high fat diet,
  obesity, and prior heart problems
• Incidence 1 in 100, typically male, starting
  with middle age
• Symptoms include chest pain, shortness of
  breath, irregular/rapid pulse, and sensation of
  feeling the heart beat
• Treatment Regimens ACE inhibitors, beta
  blockers, angioplasty (to improve blood flow to
  the damaged or weakened heart muscle), and heart
  transplant (severe cases)

7
The Basic Scientific Question

• What kinds of changes occur in cardiac
  transcription profiles brought about by heart
  failure?
• 2 ways to go about attacking the question
  Molecular Biology (hypothesis based) vs High
  Thru-put techniques (i.e. microarrays followed by
  confirmation of gene expression with qPCR)

8
Differential Expression between ischemic and
non-ischemic cardiomyopathy patients

• Gene expression analysis of ischemic and
  nonischemic cardiomyopathy shared and distinct
  gene in the development of heart failure
• M. Kittleson, K. Minhas, R. Irizarry, S. Ye,
  G. Edness, E. Breton, J. Conte, G. Tamselli, J.
  Garcia, and J. Hare. Physiol. Genomics,
  21299-307, 2005

9
Methods of Kittleson et al

• 31 cardiomyopathy vs. 6 normal patients (clinical
  characteristics were reasonably similar within
  groups)
• Tissue taken from cardio-myopathy patients at the
  time of LVAD or cardiac transplantation
• Identified differentially expressed genes in 2
  comparisons NICM (hypertrophic, valvular,
  alcholic) vs NF hearts and ICM vs NF using
  significance analysis of microarrays
• Identified genes with FDR lt 5 and absolute fold
  change greater than 2.0

10
Conclusions of Kittleson et al

• No hypothesis, but the microarray experiment was
  used to generate hypothesis
• Types of genes differentially expressed (41
  total) cell growth maintenance(9), signal
  transduction(7), metabolism(3), cell
  adhesion/cell communication(2), binding(2), and
  catalytic activity(2), nucleus(3), other (13)

11
Conclusions of Kittleson et al

• Predominance of fatty acid metabolic genes
  genesis of NICM might be metabolic in nature
• Predominance of abnormalities in catalytic
  activity with ICM (serine proteinase inhibitors)
• TNFRSF11B (member of TNF receptor subfamily) is
  significantly downregulated in ICM

12
Experimental Procedure for Data that We are Using

• Collected myocardial samples from patients
  undergoing cardiac transplantation whose failure
  arises from ischemic cardiomyopathy and from
  "normal" organ donors whose hearts cannot be used
  for transplants
• The transcriptional profile of the mRNA in these
  samples was measured with gene array technology.
• Changes in transcriptional profiles can be
  correlated with the physiologic profile of
  heart-failure hearts acquired at the time of
  transplantation.

13
Working Hypothesis?

• Because of the results of Kittleson et al, we can
  generate a simple working hypothesis
• Our differentially expressed genes, using our
  methods of statistical analysis of the data,
  should roughly be the same as what Kittleson et
  al obtained in their paper.

14
Exploratory Analysis of Data

• Goal identify genes whose expression levels
  are
• differentially expressed between Ischemic and
• Normal.
• Affymetrix Data with Two Population
• 54,675 genes are expressed
• for 32
  Ischemic samples
• 14 Normal
  samples
• How do we compare?

15

• Pre-processing
• Only obtain the expression measurement of the
  data (ie., put it into exprSet) using the
  default of justRMA method
• bgcorrect.method rma
• normalized.method quantiles
• summary.method liwong

16

• Histogram of Ischemic/Normal
• The distribution is skewed right.
• The range is between 4 to 14.
• Both histograms have similar shapes.
• Boxplot of Ischemic/Normal
• There are many outliers from the upper
  values.
• The intensity of Ischemic is higher
  than Normal.
• Histogram of MAD (Median Absolute Deviation)
• Cut-off Method by MAD
• Apply MAD gt 0.1.
• We can filter out 675 genes from a
  total of 54675 genes.
• Quantile-Quantile plot
• A visual aid for identifying genes with
  unusual test
• statistics.
• It shows the large deviation at the right
  tail.

17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22

• t-Test for
• Mean difference between Ischemic and Normal
• H0 H1
  
• We are testing 54675 genes simultaneously and
  adjust for multiple testing when assessing the
  statistical significance of the observed
  associations to control the false positive rate.

23
Multiple Hypothesis Testing

• Motivation
• To identify as many differentially expressed
  genes as possible, while incurring a relatively
  low proportion of false positives.
• H0 No differential gene expression (between
  Ischemic and normal group)
• Large multiplicity problem more than fifty
  thousand hypotheses are tested simultaneously.
• How can we control the false positive rate
  genomewide? FDR or pFDR.

24
Table1. Possible outcomes from thresholding m
genes for significance (m p-values with some
cutoff point applied).
Called significant (reject H0) Called not significant (accept H0) Total
True null (H0 is true) F ( of false positives) m0 - F m0
True alternative (Ha is true) T ( of true positives) m1 - F m1
Total S ( of sign. features) m - S m
25
False Discovery Rate

• FDR E(F/S)
• In case S0, defined to be E(F/SSgt0)P(Sgt0) or
  define F/S0 if S0.
• Alternatively, define pFDRE(F/SSgt0). When m is
  large, P(Sgt0) is approx. 1 and FDR is approx.
  equal to pFDR.
• FDR is a measure of the overall accuracy of a set
  of significant features.

26
Linear Step-Up Procedure
27
Steps

• Select desired limit q on E(FDR)

28
FDR Adjusted P-Values

• For an individual hypothesis,

29
Data inter-dependencies

• Between genes
• Between measurement errors of expression levels

- co-regulation - spatial effects

• RNA source
• normalization process
• pooled variability estimation

Multiple testing of such data will produce
correlated test statistics !
30
Correlated Test Statistics
Positive Dependency (Benjamini Yekutieli, 2001
and Yekutieli, 2002).

• The linear step-up procedure controls the FDR for
  positive dependent test statistics.

• This condition is satisfied by

- positively correlated one-sided normal and t
test statistics.
- absolute values of normal and t test
statistics, when all null hypotheses are true.
31
BH and BY procedure

• BH
• adjusted p-values for the Benjamini Hochberg
  (1995) step-up FDR controlling procedure
  (independent and positive regression dependent
  test statistics).
• BY
• adjusted p-values for the Benjamini Yekutieli
  (2001) step-up FDR controlling procedure (general
  dependency structures).

32
Our Results

• rawp BH BY
• 0 17577 17577 17577
• 1e-04 38400 37960 35207
• 2e-04 39239 38833 35935
• 3e-04 39717 39334 36373
• 4e-04 40053 39690 36714
• 5e-04 40321 39972 36966
• 6e-04 40565 40174 37166
• 7e-04 40786 40389 37370
• 8e-04 40948 40569 37513
• 9e-04 41096 40739 37661
• 0.01 41226 40885 37781

33
Plot of sorted adjusted p-values
34
Plot of adjusted p-values vs. test statistics
35
Gene Selection Analysis

• Further select genes based on the fold change
  between two conditions (Ischemic vs. Normal)
• The fold change for each gene is calculated as
  the average expression over all Ischemic samples
  divided by the average expression over all normal
  samples.

36

37
Fold change cutoff value

• There are 1495 genes with Log2(fold change) gt 1,
  and 26 genes with Log2(fold change) lt -1
• There are only 43 genes with Log2(fold change) gt
  2, and 3 genes with Log2(fold change) lt -2
• We choose the first option

38
(No Transcript)
39
Discussion

• Among the 54,675 mRNA transcripts present on the
  Affymetrix microarray platform, 675 housekeeping
  genes were filtered out.
• By selecting the adjusted P-value less than
  0.0001, only 35,207 genes are left for the
  analysis of fold change.
• After fold change selection, only 1521 genes are
  left for further selection.
• Finally, 74 up-regulated genes and 26
  down-regulated genes are selected from the
  microarray analysis for further biological
  verification and study.

40
Summary of the Selected Genes

• Of the 100 genes, there are 53 genes that have
  known biological functions. The functions of the
  other 47 genes are unknown.

41
Gene Classification

• Based on the biological process of the genes, the
  100 genes can be classified in several
  categories.

42
Biological Function Classification
43
(No Transcript)
44
(No Transcript)
45
(No Transcript)
46
Differentially Expressed Genes to ISC-Normal
Comparisons

• Among the 100 genes that are differentially
  expressed between ischemic and normal, the
  majority fell into cell adhesion, cell growth and
  maintenance, signal transduction, muscle
  contraction and development, immune response and
  regulation of transcription.
• Most of the genes are up-regulated in above
  process except one or two genes in the process of
  cell growth and maintenance and cell adhesion.
• Few genes belong to metabolism, inflammatory
  response, acute phase response and oncogenesis.

47
An important gene for Ischemic Cardiomyopathy

• Serine proteinase inhibitors has an anti-ischemic
  protective effect and has been previously
  observed in pigs subject to experimentally
  induced myocardial ischemia (Khan 2004)
  Aprotinin reduces reperfusion injury after
  regional ischemia and cardioplegic arrest.
  Protease inhibition may represent a molecular
  strategy to prevent postoperative myocardial
  injury after surgical revascularization with
  cardiopulmonary bypass.
• It was hypothesized to ben an important gene in
  Kittlesons paper (Physiol. Genomics, 2004).

48
The significance of the results

• The gene differentiation analysis find out the
  genes that either up-regulated or down-regulated
  in ischemic patients, which can correlated with
  clinical parameters in heart failure patients and
  supported ongoing efforts to incorporate
  expression profiling-based biomarkers in
  determining prognosis and response to therapy in
  heart failure.

49
Comparison with Kittleson et. al.s Paper

• Although only one common gene is found in the
  analysis, it is consistent considering the sample
  size difference, the tissue difference and the
  statistical analysis method difference.
• However, most of the genes identified from the
  analysis fell in the same categories of the
  biological functions.

50
Limitation

• Because circumstances causing a donor heart to be
  ineligible for cardiac transplantation, such as
  infection or prolonged hypotension, can also
  affect gene expression, a normal functional
  unused donor heart is not the same as a normal
  heart.

51
Future Work

• First, the gene expression profile of these 100
  genes need to be verified by the Northern Blot or
  Real-Time RT-PCR (qPCR).
• After verification, some high fold change unknown
  function genes can be chosen to study their
  functions for biologists.

52
Acknowledgements

• MBI (Prof. Friedman and staff)
• Professors Shili Lin and Joseph Verducci
• Dr. Zailong Wang
• Dr. Nusrat Rabbee