Transcript Document

Bayesian Risk Analysis
Workshop Questions
Shuji Ogino, M.D., Ph.D.
AMP Training and Education Committee
Brigham and Women’s Hospital
Dana-Farber Cancer Institute
Harvard Medical School
I would appreciate any feedback
[email protected]
Special Thanks!
• Rob Wilson, Pam Flodman, Pam Hawley,
Bert Gold and Wayne Grody; collaborators
of risk analysis projects
• Jean Amos Wilson, Vicky Pratt and
Ed
Highsmith; helpful suggestions
• AMP Program Committee, Training and
Education Committee, and Genetics
Subdivision
• Early-birds, thank you!
Q1. Cystic fibrosis
Non-Hispanic Caucasian family
Carrier screening negative
for the ACMG 23 mutation panel
CF
Sensitivity = mutation detection rate = 88%
Specificity = 100%
Hints
• Sensitivity = positive / all carriers (or
patients); you want them to be positive
• Specificity = negative / all non-carriers (or
controls); you want them to be negative
Hints
• For alternative possibilities (e.g., carrier vs.
non-carrier), always assume one of them is
true (e.g., she is a carrier), then calculate the
probability that the test result (negative)
happens (= conditional probability)
• Assuming = key to success in Bayes
Q2. Cystic fibrosis testing
Same mutation panel
Negative
for the same
23 mutation panel
Carrier risk?
Classic CF patient
Tested for the ACMG
23 mutation panel
Only one p.F508del detected
p.F508del constitutes 72% of all disease alleles
Hints
• Be careful when test result on
proband is available
• Especially watch for undetectable
mutation !
Q3. Tough but common question. Cystic fibrosis testing
Different mutation panels
Negative for an expanded
mutation panel that detects
93% of Non-Hispanic
Caucasian disease alleles
Classic CF patient
Tested for the ACMG
23 mutation panel (that detects
88% of all disease alleles)
Only one p.F508del detected
Carrier risk?
Hints
• What is the probability that
mutations undetectable by the
23-mutation panel can be
detected by the expanded panel?
Q4. Cystic fibrosis testing
Only one detectable mutation
(Almost imaginary scenario, but prelude to Q5)
Non-Hispanic Caucasian family
Only one p.F508del detected by
prenatal testing with ACMG 23-panel
(that detects 88% of disease alleles).
p.F508del constitutes 72% of all
disease alleles
What is CF disease risk?
Hints
• Assume AFFECTED fetus, then calculate
conditional probabilities
• Assume CARRIER fetus, then calculate
conditional probabilities
• Assume NON-CARRIER fetus, then
calculate conditional probabilities
For further reading: Ogino et al. J Med Genet 2004;41:e70.
Q5. Cystic fibrosis testing
Only one detectable mutation
Non-Hispanic Caucasian family
Carrier screening shows
p.F508del by the ACMG 23
mutation panel
Cond. prob. of EB
if affected = 0.11
if a carrier = 0.00089
if a non-carrier =
0.00035
Fetal echogenic bowel (EB)+
Only one p.F508del by prenatal
testing (ACMG 23 mutation panel
that detects 88% of disease alleles).
p.F508del = 72% of all mutant
alleles
What is CF disease risk?
Hints
• Start Bayesian analysis from the top
• Assume carrier father, then calculate
conditional probabilities
• Assume non-carrier father, then
calculate conditional probabilities
– For further reading: Ogino et al. J Med Genet
2004;41:e70.
From here: Advanced questions
for other diseases
Three reasons to do:
1. You can have fun in an airplane to the meeting
2. We can go over if time allows at the workshop
(answers will be available)
3. I am always happy to discuss
Q6. Autosomal Dominant Disease
Affected
Unaffected. Carrier risk?
II-2
Age 65
Penetrance at age 65 = 0.6
at age 40 = 0.3
Unaffected.
Carrier risk?
III-1
Age 40
Hints
• One comprehensive Bayesian table
gives all correct answers at once
• Information of a child may modify
risks of the parents and
grandparents
Q7. Autosomal Dominant Disease
with Age-dependent Penetrance
Affected
Unaffected at age 50
Heterozygous risk?
Disease risk by age 70?
Penetrance by age 50 = 0.4
by age 70 = 0.75
Hints
• If he is a carrier, what is the risk to
become symptomatic from age 50 to
70?
Q8. Consanguinity: IV-1’s risk for rare AR disease
I-1
II-1
I-2
II-2
II-3
III-2
III-1
IV-1
II-4
III-3
III-4
IV-2
V-1
Carrier
AR disease
Hints
• Watch for dependent possibilities
• Start from a key person (connector
between consanguineous couple and
proband).
Q9. Isolated Case of X-linked Recessive Disease
I-2
II-1
Carrier
Risk?
II-2
Carrier
Risk?
II-3
II-4
Carrier
Risk?
Age 30, 36 asymptomatic
Age 2 months
DMD
Assume  =  (same maternal and paternal de novo mutation rate)
Hints
• Start analysis from the top
• One comprehensive Bayesian table
can give all answers at once
• Carrier risk = 4 (given  = ) for a
woman with no relative affected with
lethal X-liked recessive disease