A simple method to localise pleiotropic QTL using univariate linkage analyses of correlated traits

1
A simple method to localise pleiotropic QTL using
univariate linkage analyses of correlated traits
Manuel Ferreira Peter Visscher Nick Martin David
Duffy
2
A simple method to identify pleiotropic QTL
using univariate association analyses of
correlated traits
Manuel Ferreira Peter Visscher Nick Martin David
Duffy
3
Background
4
Localize regions of the genome that may harbour
susceptibility loci for a complex trait (e.g.
asthma, obesity, schizophrenia)
Multiple intermediate (related) phenotypes (e.g.
IgE, lung function)
Linkage analysis of multiple traits
5
Traits
Marker
T1
T2
T3
T4
T5
T6
T7
Observed LOD scores
Q
0.8 , 2.2 , 2.8 , 3.7 , 2.1 , 0.7 , 0.0
1. Significance of linkage between one marker and
one trait
Ferreira et al. 2005 Am J Hum Genet, in press
2. Significance of linkage between one marker and
multiple traits
Ferreira et al. Eur J Hum Genet, submitted
Is the observed linkage between multiple traits
and Q stochastic or a result of an underlying
pleiotropic QTL or multiple clustered QTL?
Multivariate VC/HE, PC
6
1. Linkage between a marker and multiple traits
Combined-sum
7
Combined-sum approach
Traits
Marker
T1
T2
T3
T4
T5
T6
T7
1.
Vector of LOD scores
Observed
Q
0.8 , 2.2 , 2.8 , 3.7 , 2.1 , 0.7 , 0.0
Simulated
R1
2.4 , 0.1 , 3.2 , 0.0 , 1.2 , 0.4 , 0.6
R2
1.8 , 0.3 , 0.1 , 2.7 , 1.6 , 0.0 , 0.0

R10,000
4.1 , 0.0 , 0.0 , 1.4 , 0.2 , 1.6 , 0.4
2.
Vector of empirical pointwise P-values
(-log10P)
Q
Observed
1.0 , 2.6 , 3.2 , 4.0 , 1.3 , 0.4 , 0.3
R1
Simulated
2.0 , 0.4 , 3.6 , 0.3 , 0.8 , 0.3 , 0.9
R2
1.6 , 0.5 , 0.4 , 3.1 , 1.0 , 0.3 , 0.3

R10,000
4.4 , 0.3 , 0.3 , 2.4 , 0.3 , 1.2 , 0.6
8
Combined-sum approach
3.
Vector of ordered empirical pointwise P-values
(-log10P)
Q
4.0 , 3.2 , 2.6 , 1.3 , 1.0 , 0.4 , 0.3
Observed
R1
Simulated
3.6 , 2.0 , 0.9 , 0.8 , 0.4 , 0.3 , 0.3
R2
3.1 , 1.6 , 1.0 , 0.5 , 0.4 , 0.3 , 0.3

R10,000
4.4 , 2.4 , 1.2 , 0.6 , 0.3 , 0.3 , 0.3
Sum statistics
S1
S2
S3
S4
S5
S6
S7
4.
Compute m sum statistics, Sk
Observed
Q
4.0 7.2 9.8 11.1 12.1
12.5 12.8
Simulated
R1
3.6 5.6 6.5 7.3 7.7
8.0 8.3
R2
3.1 4.7 5.7 6.2 6.6
6.9 7.2

R10,000
4.4 6.8 8.0 8.6 8.9
9.2 9.5
9
Combined-sum approach
Sum statistics
S1
S2
S3
S4
S5
S6
S7
5.
Determine significance of each Sk
Observed
Q
.045 .023 .015 .012 .007
.007 .007
Simulated
R1
.065 .055 .050 .045 .045
.045 .047
R2
.150 .120 .101 .101 .102
.102 .104

R10,000
.035 .030 .025 .020 .020
.021 .025
Hoh et al. 2001 Genome Res 11 2115-2119.
Dudbridge Koeleman 2004 Am J Hum Genet 75
424-435.
Observed
Q
.045 .023 .015 .012 .007
.007 .007
6.
Identify the Sk with smallest P-value and assess
its significance
Simulated
R1
.065 .055 .050 .045 .045
.045 .047
R2
.150 .120 .101 .101 .102
.102 .104

R10,000
.035 .030 .025 .020 .020
.021 .025
Global empirical pointwise P .011
e.g. S5(Q) .007
10
Simulations
Power Combined-sum
3 traits 1 QTL for 250 sib-pairs
Eight models varied QTL variance trait
correlation
1,000 datasets
11
Simulations
If multiple correlated traits are analysed, power
to detect a QTL can be improved by considering
all traits simultaneously
Combined-sum approach is an efficient alternative
to formal multivariate methods, applicable to any
number of traits not affected residual
correlation
12
2. Application asthma dataset
13
Application to asthma
215 sib-pairs (201 families)
Measured for 7 asthma traits Asthma, BHR, Atopy,
Dpter, FEV1, FEV1/FVC, IgE
6 201-300 markers, 48 301-500, 25 501-1000 and
21 1001-1544
Information content 0.57 (range 0.15-0.85)
Significance estimated 1,796,000 (univariate)
marker replicates
14
Application to asthma II
Mixture continuous affection traits
Dpter Atopy BHR FEV1 Asthma FEV1/FVC IgE
15
Application to asthma II
Global genome-wide P .023
16
Conclusion
1. Developed an efficient approach to test
whether the simultaneous linkage of multiple
traits to the same marker is spurious
All traits must be analysed with the same marker
replicates generated under the null hypothesis of
no linkage
2. This approach is more powerful than univariate
VC to detect a pleiotropic QTL
3. When traits are moderately correlated and the
QTL influences all traits it outperforms
multivariate VC
It is applicable to any number of traits and it
is not affected by the residual correlation
between traits
4. Further testing required to assess performance
under specific situations
Longitudinal data, many traits, different linkage
statistics
5. Applicable to association analysis, including
genome-wide
17
Acknowledgments
David Smyth
Allan McRae
Carl Anderson