Transcript PowerPoint-Präsentation

1
Masterseminar
„A statistical framework
for the diagnostic of
meningioma cancer“
Andreas Keller
Supervised by: Professor Doktor H. P. Lenhof
Chair for Bioinformatics,
Saarland University
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
2
Introduction
What are
meningiomas
 Benign brain
tumors
 Arising from
coverings of brain
and spinal cord
 Slow growing
 Most common
neoplasm (brain)
 Genetic alterations
3
Introduction
4
Introduction
meningioma in proportions
Two times more often
in women as in men
More often in people
older than 50 years
5
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
6
SEREX
se
serological identification
of antigens
by rrecombinant ex
expression cloning
7
SEREX – Identification
expression of a human
fetal brain library
proteins bind
on membrane
2nd antibody
detection
8
pooled sera
SEREX – Screening
agar plate
patients serum
9
specific genes
2nd antibody
detection
SEREX – Results
10
Microarrays
System:
cDNA microarrays
55.000 spots
Whole Genome Array
Data:
8 samples per WHO grade
2 dura as negative controle
2 refPools as negative controle
11
Microarrays
12
Statistical Learning
Supervised Learning
Bayesian Statistics
Support Vector Machines
Discriminant Analysis
Unsupervised Learning (Clustering)
Feature Subset Selection
Component Analysis (PCA, ICA)
13
Statistical Learning
Crossvalidation
Error Rates
Training Error
CV Error
Test Error
Specificity vs. Sensitivity tradeoff
Receiver Operating Caracteristic Curve
14
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
15
SEREX
Data situation:
p = 57
n = 104
Goal:
Predict meningioma vs. non meningioma
Predict WHO grade
16
Bayesian Approach
serum 1
serum 2
serum 3
serum 4
serum 5
serum 6
serum 7
serum 8
serum 9
serum 10
serum 11
serum 12
class
gene A
gene B
1
1
2
2
3
3
0
0
0
0
0
0
0
1
1
1
0
1
0
0
1
0
0
0
1
1
1
0
1
0
0
0
0
0
0
0
17
Bayesian Approach
serum 1
serum 2
serum 3
serum 4
serum 5
serum 6
serum 7
serum 8
serum 9
serum 10
serum 11
serum 12
class
gene A
gene B
1
1
2
2
3
3
0
0
0
0
0
0
0
1
1
1
0
1
0
0
1
0
0
0
1
1
1
0
1
0
0
0
0
0
0
0
4
6
1
6
4
6
1
0
7
6
18
Bayesian Approach
19
Bayesian Approach
serum 1
serum 2
serum 3
serum 4
serum 5
serum 6
serum 7
serum 8
serum 9
serum 10
serum 11
serum 12
class
gene A
gene B
1
1
2
2
3
3
0
0
0
0
0
0
0
1
1
1
0
1
0
0
1
0
0
0
1
1
1
0
1
0
0
0
0
0
0
0
2
6
5
6
2
6
6
6
20
Bayesian Approach
21
Bayesian Approach
22
SEREX Conclusion
Separation meningioma vs. non
meningioma seems very well possible
Separation into different WHO grades
seems to be possible with a certain
error
23
SEREX Conclusion
Extend to other
Brain tumors (glioma)
Human cancer
Disease
Simplify experimental methods
Develop a prediction system
24
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
25
Microarray
Data situation:
p = 53423
n = 26
2 goals:
Find significant genes
Classify into WHO grades
26
Dimension reduction
6 approaches
Component analysis
Take genes which differ from DURA
Take genes which differ from refPool
Take genes which differ between grades
Take „publicated“ genes
Split into chromosomes
27
Component analysis
Principal component analysis
Independant component analysis
28
Analysis of grades
tissues
genes
29
Dura and refPool
Justification for Dura
 Wherefrom to take?
 How to take?
 Genes different from normal tissue
 Good to classify into meningioma vs. healthy
Justification for refPool
 Genes different between WHO grades
 Good to classify into grades
30
Published genes
Several 100 genes are connected with
meningioma in several publications
Find these genes and investigate them
example: Lichter 2004 – 61 genes with different
expression WHOI in contrast to WHOII and III
31
Split into chromosomes
As mentioned: often karyotypic alterations
losses:
gains:
 22
 1p
 6q
 10q
 14q
 18q
 1p
 9q
 12q
 15q
 17q
 20q
=> Split genes into different chromosomes
=> Compare to karyotype
32
Split into chromosomes
33
Classification
Classification:
Clustering
SVM
Discriminant Analysis
Least Squares
34
SEREX derived genes
35
BN++
BN++ as a statistical tool
Build a C++/R interface??
Use MatLab??
Use C++ librarys??
36
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
37
Workflow
Large scale investigation of
suspicious people by antigen
analysis.
If a positive prediction is made do
further analysis (CT or similar).
If necessary surgory.
Further examinations with the
gained tissue.
38
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
39
Outline
Outline
Introduction
Materials and Methods
SEREX
Microarray
Conclusion
Discussion
40