Академический Документы
Профессиональный Документы
Культура Документы
Two
addi8ve
(non-epista8c)
loci
Dominant epistasis
One
locus
in
a
dominant
way
suppresses
the
allelic
eects
of
a
second
locus
Epistasis Example
Co-adap8ve epistasis
Almost
no
marginal
eects:
average
eect
of
JJ,
JL,
LL
do
not
dier
Epistasis Example
Dominance-by-
dominance
epistasis
Double
heterozygote
(LS,
LS)
deviates
from
the
phenotype
that
is
expected
from
the
phenotypes
of
the
other
heterozygotes.
Double
heterozygotes
have
a
lower
phenotype
than
expected.
Epistasis
Use B-sta8s8c
p1=p2
p1=p2
=0.0033
=0.001
p1=p2
p1=p2
=0.0001
=0.00001
Simulated
data
AMD
data
Screen and Clean (Wu et al., 2010)
Assume
that
the
SNPs
with
epista8c
eects
are
likely
to
have
at
least
small
marginal
eect
Select
the
SNPs
with
marginal
eects.
Consider
only
those
SNPs
with
marginal
eects
for
epista8c
interac8on.
SNPs
will
be
found
to
have
epista8c
eects
but
no
marginal
eects
Screen: Two-stage Lasso for Detecting
Epistasis
Two-stage
lasso
Step
1:
Apply
lasso
with
no
considera8on
of
epistasis
to
detect
SNPs
with
signicant
individual
eects
Step
2:
Apply
lasso
with
pairs
of
only
those
SNPs
selected
in
Step
1
Clean: Assessing Significance of Results
Mul8-split
method
Randomly
split
the
data
into
Stage
1
and
Stage
2
datasets
mul8ple
8mes
Perform
the
two-way
split
method
for
each
split
Combine
the
p-values
from
each
split.
Advantages and Disadvantages
Need
to
split
the
data
into
two
parts
for
Stage
1
and
Stage
2
analysis
Screen and Clean
Simulation Results
WTCCC
data
1,963
cases
with
T1D
and
2,938
controls
Mul8-split
screen
and
clean
with
56
random
splits