Title: Summary of Graphical and Causal Modeling in Genetics and Epidemiology by Vanessa Didelez Nuala Sheeh
1Summary ofGraphical and Causal Modeling in
Genetics and Epidemiologyby Vanessa
DidelezNuala Sheehan
- Hein Stigum
- http//folk.uio.no/heins/
Nov-09
H.S.
1
2Graphical Models
- Pedigrees
- Compute carrier probability
- Causal Reasoning
- Conditional independence
- Two main types
- Undirected graphs
- Directed acyclic graphs
- Medelian randomization
Nov-09
H.S.
2
3Introduction
Nov-09
H.S.
3
4Associations
- E and D associated if
- E causes D
- C is a common cause of E and D
- C is a common effect of E and D and we condition
on C -
Confounder
Collider
Nov-09
H.S.
4
Nov-09
H.S.
4
5Marginal and conditional dependence
Conditionally Independent C
Marginally dependent
E and D are
Nov-09
H.S.
5
6Test of H0
- Can test H0 only if
- E and D are conditionally independent, given the
variables we adjust for (C)
Not OK
OK
Need graphic tools!
Nov-09
H.S.
6
7Graphic tools
Nov-09
H.S.
7
8Undirected Graph
Definition A and B are separated by C if all
paths from A to B pass thru C
Are 1 and 5 separated by 2 ? Are 1 and 5
separated by 3 ? Are 1 and 5 separated by 3 and 4
?
Yes No Yes
If 1 and 5 are separated by 2, then 1 and 5 are
conditionally independent given 2
Nov-09
H.S.
8
9Example Chewing tobacco and ulcers
T and U are marginally dependent
T and U are conditionally independent given A
T and U are separated by A
Nov-09
H.S.
9
10Directed Acyclic Graphs, DAGs
1
3
5
Are 1 and 4 separated by 2 ?
No
2
4
- Steps
- Take ancestral graph of 1,2,4
- Moralize the graph
- Look for separation
Nov-09
H.S.
10
11Directed Acyclic Graphs, DAGs
1
3
5
Are 1 and 4 separated by 2 ?
No
2
4
1
3
Take ancestral graph of 1,2,4
2
4
1
3
Moralize the graph
2
4
Are 1 and 4 separated by 2 ?
No
Nov-09
H.S.
11
12Estrogen and Endometrial cancer
Uunknown uterine abnormality Dendometrial
cancer Eestrogen Cvaginal bleeding Aascertained
cancer
U
D
A
E
C
Null hypothesis E and D independent. Can we test
H0? Case-control study, condition on A. Are E
and D cond. independent given A? Does it help to
adjust for C?
1. Take ancestral graph of E,D,A 2. Moralize 3.
Separation by A? 4. Separation by C,A?
No
No
Nov-09
H.S.
12
13Test of H0
- Can test H0 only if
- E and D are conditionally independent, given the
variables we adjust for (C)
Not OK
OK
Use tools to verify!
Nov-09
H.S.
13
14Cause versus Association
- Observe association, not necessarily causal
- Confounding
- Reverse causation
- Selection effect
- Time trends
Nov-09
H.S.
14
15Definitions
- Association
- Observing E predicts D
- Causation
- Manipulating E predicts D
Nov-09
H.S.
15
16Problem
- Association ? causation
- Intervention on association may be useless
- Randomization not always feasible
- Need causal information from observational studies
Nov-09
H.S.
16
17Mendelian randomization
Nov-09
H.S.
17
18Observation versus trial
- Observational study
- all measured confounders adjusted for
- Randomized trial
- Strong effect, compliance
- Does not exist
- Does not exist
- R?D if and only if E causes D
- Medelian randomization
- Gene/ exposure association strong, or large N
- Should not exist, Mendels 2. law
- Must not exist, depends on the function of the
gene - G?D if and only if E causes D
U
2
E
D
R
1
3
U
2
E
D
G
1
3
Nov-09
H.S.
18
19Ex Alcohol and blood pressure
- Observational study
- Alcohol use increases blood pressure
- Many lifestyle confounders
- Gene ALDH2, 2 alleles
- 2,2 type suffer nausea, headache after alcohol
- ? low alcohol regardless of lifestyle (U)
- Medelian randomization
- Gene ALDH2 is highly associated with alcohol
- Mendels 2. law, no ass. to obs. confounders
- OK, gene function is known
- Result 2,2 type BP 7.4 mmHg
U
2
A
BP
G
1
3
Nov-09
H.S.
19
20Violations of core conditions
- Gene/exposure association
- Gene rare or weak effect ? large N
- Compensation
- and 3. Gene independent of U and D
- Pleiotropy
- Linkage disequilibrium
- Population stratification