HAL Id: hal-00421215
https://hal.archives-ouvertes.fr/hal-00421215
Preprint submitted on 1 Oct 2009
HAL
is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire
HAL, estdestinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Likelihood Ratio Test process for Quantitative Trait Loci detection.
Charles-Elie Rabier, Jean-Marc Azaïs, Céline Delmas
To cite this version:
Charles-Elie Rabier, Jean-Marc Azaïs, Céline Delmas. Likelihood Ratio Test process for Quantitative
Trait Loci detection.. 2009. �hal-00421215�
Likelihood Ratio Test process
for Quantitative Trait Loci detection
Charles-Elie Rabier
Institut de Mathématiques de Toulouse, Toulouse, France.
INRA UR631, Auzeville, France.
Jean-Marc Azaïs
Institut de Mathématiques de Toulouse, Toulouse, France.
Céline Delmas
INRA UR631, Auzeville, France.
Summary. We consider the likelihood ratio test (LRT) process related to the test of the ab- sence of QTL on the interval[0, T]representing a chromosome (a QTL denotes a quantitative trait locus, i.e. a gene with quantitative effect on a trait). We give the asymptotic distribution of this LRT process under the null hypothesis that there is no QTL on[0, T]and under the general alternative that there existmQTL on[0, T]. We propose to estimate the number of QTL, their positions and their effects by penalized likelihood. Our results are extended to the case where individuals are structured into families.
Keywords: Gaussian process, Likelihood Ratio Test, Mixture models, Nuisance parameters present only under the alternative, QTL detection,χ2process.
1. Introduction
n
Yj, j= 1, ..., n
! "
# $ A×(A×B) AB % &
'!(!(
[0, T] [0, T])
$ t=t Y
* $ +
p(t)f(µ+q,σ)(.) +{1−p(t)}f(µ−q,σ)(.) !
f(µ,σ)(.) , µ σ2 (µ, q, σ)
$ - t ∈ [0, T] $
. “q = 0” ! n Y1, ..., Yn
Λn(t)
/ 0 pj(t) p(t) 1
{Λn(t), t∈[0, T]} 2$ 3 $
* .
*
4 5 !
6 !(7( . [0, T]
8&$ 9
!((7 6 * % 8
&$ -%:9 - "55"-%:
"55( / .
.: !((;
<!(==<!(7= *
>,
4
4 ) ? .
[0, T] [0, T]q= 0
t [0, T]
* !t=t
.
*m [0, T]t1,· · · , tm q1,· · · , qm
* M = 2m + M
α=1
pαf(mα,σ)(.)
pαmα$ $ µmt1tm q1qm &.
, $
$ % $
6 . 2
3 .
*
# $ .
.:!((;.:!((@ 2
3 % ?
*
. .
9 "55( 8 $
*7@
?
.: *
% $
$
AA!((7 4
2. Model
[0, T] K $
* t1 = 0 < t2 < ... < tK =T $
2 3t X(t) '!(!(
+ N(t) B X(t) 12 (δ1+δ−1)
X(t) = (−1)N(t)X(t1) '!(!(C r: [0, T]2−→
0, 12
+
r(t, t) =P(X(t)X(t) =−1) =P(|N(t)−N(t)| ) = 1
2 (1−e−2|t−t|)
¯
r(t, t) 1−r(t, t)
Y X(t)t ∈[t1, tK]
D+
Yj =µ + X(t)q + σε
ε, q
6 2 3 $
t1, t2, ..., tK Xj(t) X(t) 1
/ (Yj, Xj(t1), ..., Xj(tK))
% t $ 3. Only 2 genetic markers
$(K= 2) 0T + 0 =t1<
t2=T -* $ t ∈[t1, t2]
t ∈[t1, t2] 4 p(t) Dp(t) =P
X(t) = 1X(t1), X(t2)
9 *X(t1) =X(t2) = 16+
P
X(t) = 1X(t1) = 1, X(t2) = 1
= (1/2)P{N(t)−N(t1)}P{N(t2)−N(t)} (1/2)P{N(t2)−N(t1)}
/ ∀t∈]t1, t2[+
p(t) =Q1t,11X(t1)=11X(t2)=1 + Q1t,−11X(t1)=11X(t2)=−1
+Q−t1,11X(t1)=−11X(t2)=1 + Q−t1,−11X(t1)=−11X(t2)=−1 "
+
Q1t,1= ¯r(t1, t) ¯r(t, t2)
¯
r(t1, t2) , Q1t,−1= r(t¯ 1, t)r(t, t2) r(t1, t2) Q−t1,1= r(t1, t) ¯r(t, t2)
r(t1, t2) , Q−t1,−1= r(t1, t)r(t, t2)
¯ r(t1, t2)
$+
Q−t1,−1= 1−Q1t,1 Q−t1,1= 1−Q1t,−1
6 p(t1) = 1X(t1)=1 p(t2) = 1X(t2)=1 / p(t) t1
t2
θ= (q, µ, σ) tD*θ0= (0, µ, σ)
H0 $ (Y, X(t1), X(t2))
λ⊗N ⊗N λ N N
∀t∈[t1, t2]+
L(θ, t) =
p(t)f(µ+q,σ)(y) +{1−p(t)}f(µ−q,σ)(y)
g(t) 0
g(t) =1 2
r(t¯ 1, t2) 1X(t1)=11X(t2)=1 + r(t1, t2) 1X(t1)=11X(t2)=−1
+1 2
r(t1, t2) 1X(t1)=−11X(t2)=1 + ¯r(t1, t2) 1X(t1)=−11X(t2)=−1
$ Ln(θ, t) n n
θˆ= (ˆq, µ,ˆ σ)ˆ *$ )E θ
&H0 [t1, t2] 6 H1
tt∈[t1, t2]
D H1 / +
Hat :2 t q=a/√
na∈R 3
q q=a/√
n 9!(7FC
3.1. Results
Sn(.)⇒Z(.) Λn(.)F.d.→ {Z(.)}2
H0 Hat
• Sn(.) n
• ⇒ F.d.→
• Z(.) ∀(t, t)∈[t1, t2]2
Γ(t, t) = 4E{p(t)p(t)} −1
E
{2p(t)−1}2 E
{2p(t)−1}2
∀(t, t)∈[t1, t2]2
• H0m(t) = 0
• Hat
mt(t) = aE[X(t){2p(t)−1}] σ
E
{2p(t)−1}2
Z(.)Z(.)
∀t∈[t1, t2]
Z(t) ={α(t)Z(t1) + β(t)Z(t2)}/
E
{2p(t)−1}2
∀t∈]t1, t2[ α(t) =Q1t,1+Q1t,−1−1 β(t) =Q1t,1−Q1t,−1α(t1) = 1β(t1) = 0 α(t2) = 0β(t2) = 1{Z(t1), Z(t2)}=e−2t2
! ∀(t, t)∈[t1, t2]2
mt(t) ={α(t)mt(t1) + β(t)mt(t2)}/
E
{2p(t)−1}2
"# E
{2p(t)−1}2 $%&'(
) % E{p(t)p(t)} * % E[X(t){2p(t)−1}] $%+'
( ) %
D
% "55=
?! Γ(t, t) mt(t) T
0.2) $
9 -%:"55F-%:"55( t Γ(t, t)
Y 2 3 X(t1) X(t2) -
) 9
Fig. 1.Mean function and Covariance function (a= 4,σ= 1,T= 0.2M)
$ .
∀t∈[t1, t2]+
Λn(t) ={Wn(t)}2+oPθ
0(1) ={Sn(t)}2+oPθ
0(1)
Wn(t)Sn(t) n
/ =! oPθ
0(1)
% H0
- ( / =!+
Wn(t) =√ nqˆ
E
{2p(t)−1}2 /σ , Sn(t) =n
j=1
(yj−µ) (2pj(t)−1)
√n σ
E
{2p(t)−1}2
;
# σ σˆ /$C
µ µˆ B σ
ˆσ /$C # $
D ;
! ' !
!
- $ +
Sn(t) ={α(t)Sn(t1) + β(t)Sn(t2)}/
E
{2p(t)−1}2 @
9
Sn0(t1), Sn0(t2)
=e−2t2 Sn0(.) H0
4 +
Λn(t) ={α(t)Sn(t1) + β(t)Sn(t2)}2/E
{2p(t)−1}2 + oPθ
0(1)
={α(t)Wn(t1) + β(t)Wn(t2)}2/E
{2p(t)−1}2 + oPθ
0(1)
6 !/ =!oPθ
0(1)
% Hat . t
$
$
# $
Λn(.)
E) )EC
$
3.2. Remarks
* Sn(.)Λn(.) Vn(.)
Sn(.)+
•
• %
) +
Vn(t) =
t2−t
t2 Sn(t1) + t
t2 Sn(t2)
/
τ(t) F
τ(t) =V t2−t
t2 Sn0(t1) + t
t2 Sn0(t2)
=
t2−t t2
2
+ 2 t(t2−t)
(t2)2 e−2t2 + t
t2
2
4 τ(t)= 0 ∀t ∈[t1, t2] Vn(.) ,
H0 9
Sn0(t1), Sn0(t2)
=e−2t2
/ Vn(.)+
- !! / =!
Vn2(.) $
* +
p(t) = 1X(t1)=11X(t2)=1 + t2−t
t2 1X(t1)=11X(t2)=−1 + t
t2 1X(t1)=−11X(t2)=1 =
$ * D
" /
* $
Vn2(.) % H0 .: !((@+
*
n¯r(t1, t2)/2 nr(t1, t2)/2 nr(t1, t2)/2 n¯r(t1, t2)/2 n
j=11Xj(t1)=11Xj(t2)=1n
j=11Xj(t1)=11Xj(t2)=−1n
j=11Xj(t1)=−11Xj(t2)=1 n
j=11Xj(t1)=−11Xj(t2)=−1
6 ! / =! Hat Vn(.)
H0 m˜t(t)
m˜t(t) +
˜ mt(t) =
t2−t
t2 mt(t1) + t
t2 mt(t2)
/ τ(t)
Vn(.) D 9
Sn0(t1), Sn0(t2)
=e−2t2 4
S0n(t1)Sn0(t2)τ(.) 4
Vn2(.)% .:
!((@ Sn0(t1)Sn0(t2) E
{2p(t)−1}2 = 0
p(t)D =
4. Several markers : the “Interval Mapping‘’ of Lander and Botstein (1989)
4 K $0 = t1 < t2 < ... < tK =T
t t t $
$ ? t ∈[t1, tK]\Tk
Tk={t1, ..., tK}Dt tr+
t=sup{tk ∈Tk:tk< t} , tr=inf{tk∈Tk :t < tk}
4 t 2)$3(t, tr)
%
E
{2p(t)−1}2 , E{p(t)p(t)} , E[X(t){2p(t)−1}] , α(t) , β(t)
,Z(.) Z(t) =
α(t)Z(t) + β(t)Z(tr) /
E
{2p(t)−1}2
∀k∀k{Z(tk), Z(tk)}=e−2|tk−tk|
! mt(t) mt(t) =
α(t)mt(t) + β(t)mt(tr) /
E
{2p(t)−1}2
* -
# ∀k ∀kΓ(tk, tk) =e−2|tk−tk| 4 8&$
6 !(7(9 !((7
6 / 0!+
∀k Wn(tk) =√
nq/σˆ , Sn(tk) = n j=1
(yj−µ) (2 1Xj(tk)=1−1) σ √
n
Λn(t) =
α(t)Sn(t) + β(t)Sn(tr)2
/E
{2p(t)−1}2 + oPθ
0(1) 7
=
α(t)Wn(t) + β(t)Wn(tr)2
/E
{2p(t)−1}2 + oPθ
0(1)
# ∀k∀k9
Sn0(tk), Sn0(tk)
=e−2|tk−tk|
6 *7"oPθ
0(1) %
Hat
4.1. Remarks
Vn(.)/ 0"%
$ % H0
.:!((; * 70
8
* X(t) $
/
8&$ $
$ Mn(.) -