Transcript Slide 1

In the Name of Allah
Seminar Title:
Gene expression modeling through positive
Boolean functions
By seyyedeh Fatemeh Molaeezadeh
Supervisor: Dr. farzad Towhidkhah
31 may 2008
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
1
Outlines
Biological concepts
Microarray Technology
Gene Expression Data
Biological characteristics of gene expression data
Modeling Objects
Modeling Issues
The Mathematical Model
An application to the evaluation of gene selection methods
Conclusions
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
2
Biological concepts
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
3
Microarray Technology
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
4
Gene Expression Data
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
5
Biological characteristics of gene
expression data
 Expression Profiles
 a collection of gene expression signatures
 Expression signatures
 a cluster of coordinately expressed genes
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
6
Characteristics of gene expression
signatures
Differential expression and co-expression
Gene expression signatures as a whole rather than single genes contain
predictive information.
Genes may belong to different gene expression signatures at the same time
Expression signatures may be independent of clinical parameters
Different gene expression profiles may share signatures and may differ only
for few signatures
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
7
Modeling Objects
Evaluation of the performance of a statistic or learning
methods such as gene selection and clustering
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
8
Modeling Issues
1.
Expression profiles may be characterized as a set of gene
expression signatures
2.
Expression signatures are interpreted in the literature as a set
of coexpressed genes
3.
the model should permit to define arbitrary signatures
4.
Genes may belong to different signatures at the same time.
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
9
The number of genes within an expression signature usually
vary from few units to few hundreds.
the model should reproduce the variation of gene expression
data.
Not all the genes within a signature may be expressed in all the
samples.
Different expression profiles may differ only for few signatures
The model should be sufficiently flexible to allow different
ways of constructing an expression profile.
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
10
The Mathematical Model
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
11
a Boolean function defined on binary strings in
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
12
cardinality
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
13
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
14
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
15
An alternative way of representing a positive Boolean function
Definition 1.
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
16
Definition 2.
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
17
For example in slide 15
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
18
Other example
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
19
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
20
An application to the evaluation of gene
selection methods
• Dataset:
– 100 artificial tissues, 60 belonging to the first class and 40 in the
second class, with 6000 virtual genes.
• Gene selection method:
– Golub method (a simple variation of the classic t-test)
– the SVM-RFE procedure
• Evaluation method:
– Intersection percent between selected gene set from above mentioned methods
and marker gene set that we produce
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
21
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
22
Conclusions
• introduce a mathematical model based on positive
Boolean functions
• take account of the specific peculiarities of gene
expression
• the biological variability viewed as a sort of random
source.
• Present an applicative example.
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
23
Reference
Francesca Ruffino; Marco Muselli; Giorgio Valentini. “Gene
expression modeling through positive Boolean functions”,
International Journal of Approximate Reasoning, Vol. 47,
2008, pp. 97–108
S. F. Molaeezadeh-31 may
2008
Gene expression modeling through positive Boolean functions
24