You are on page 1of 14

A few FAQs About QTL

Mapping
Basics of QTL Mapping
Guoyou Ye

E-mail: g.ye@cigar.org

What is QTL Mapping?


Dataset of QTL Mapping
Map individual genetic factors on the quantitative traits,
to specific chromosomal segments in the genome.  Mapping population
The key questions in QTL mapping studies are:
 Linkage map (can be constructed
 How many QTL are there?
using your own marker genotype
 Where are they in the marker map?
data)
 How large an influence does each of them have on
the trait of interest?  Marker genotype
 Phenotypic data
Mapping Population

 Biparental populations
 F2
 BC: backcross
Maximum Likelihood Method for
 DH, doubled haploids Estimating Recombination Rate
 RIL, recombination inbred lines
 CSSL: chromosome segment substitution
lines
 Natural population

Recombination Rate Expected Marker Genotype Frequencies


BC1 BC2 DH Counts Expected Freq
 Recombinant gametes: are derived by
M1M1M2M2 M1m1M2m2 M1M1M2M2 n1 -r)
f1= 12 (1-
crossing over between the two parental
gametes received by an individual. M1M1M2m2 M1m1m2m2 m1m1M2M2 n2 f2= 12 r

 Recombination fraction: is the M1m1M2M2 m1m1M2m2 m1m1M2M2 n3 f3= 12 r

probability that gamete transmitted by M1m1M2m2 m1m1m2m2 m1m1m2m2 n4 -r)


f4= 12 (1-
an individual is recombinant.
Estimating Recombination Rate
Example
 Likelihood function  BC experiment
n!  P1 and P2 are AABB and aabb, respectively
L= [ 12 (1 − r )]n1 [ 12 r ]n2 [ 12 r ]n3 [ 12 (1 − r )]n4
n1! n2 ! n3 ! n4 !  Number of plants of different genotypes in the
lnL = lnC + (n1 + n4 ) ln(1− r) + (n2 + n3 ) lnr population
 Logarithm operation of the likelihood function
n2 + n3 n + n3  AABB:162;AABb:40;AaBB:41;
rˆ = = 2 AaBb:158
n1 + n2 + n3 + n4 n
 Maximum likelihood estimate 40 + 41 81
rˆ = = = 20.20%
162 + 40 + 41 + 158 401
d 2 ln L n +n n +n n
I = −E( ) = − E[− 1 42 − 2 2 3 ] =
d 2r (1 − r ) r r (1 − r )
rˆ(1 − rˆ)
 Information: The negatives of the second derivatives Vrˆ = ≈ 4.02 × 10 −4
n
1 rˆ(1 − rˆ)
Vrˆ = =
I n

Recombination Rate When Three


Markers Are Linked Mapping Function
r13 = r12 + r23 − 2(1 − δ )r12 r23  Mapping distance
Whenδ=0 (Crossovers in the two intervals are m13 = m12 + m23
independent),  Unit of mapping distance:Morgan (M)or centi-Morgan (cM),
(1 − r13 ) = (1 − r12 )(1 − r23 ) + r12 r23
or 1M= 100 cM

r13 = r12 (1 − r23 ) + (1 − r12 )r23 = r12 + r23 − 2r12 r23  Morgan is defined as that distance along which one crossing over is
expected to occur per gamete per generation
when δ= 1 (Complete interference, crossover at on  Mapping distance m is function of recombination rate:
interval completely prevent crossover event in the other m = f (r )
interval)  f is called mapping function
r13 = r12 + r23
Common Mapping Functions
• Morgan 三种作图函数的比较
m =r ×100 (cM)
• Haldane: no interference
r = 12 (1 − e − 2 m )

• Kosambi with interference

1 + 2r 1e
m / 25
−1
m = 25 ln r=
1 − 2r 2 em / 25 + 1

Basic Principle of QTL


Mapping Maximum likelihood Estimation
 Distribution of trait values of three  A common approach in
marker genotypes of a maker statistical estimation
• Define hypotheses
• Generate likelihood function
• Estimate
• Test hypotheses
• Draw statistical conclusions
15
Expected Marker Genotype
Hypotheses in Linkage Analysis Frequencies

 H 0: BC1 BC2 DH Counts Expected Freq

 θ = 0.5 M1M1M2M2 M1m1M2m2 M1M1M2M2 n1 -r)


f1= 12 (1-

 the QTL is not linked to the marker(s)


M1M1M2m2 M1m1m2m2 m1m1M2M2 n2 f2= 12 r

M1m1M2M2 m1m1M2m2 m1m1M2M2 n3 f3= 12 r

 HA: M1m1M2m2 m1m1m2m2 m1m1m2m2 n4 -r)


f4= 12 (1-
 θ ≠ 0.5
 the QTL is linked to the marker(s)

QTL Effects and QTL Genotypic


Value :What is LOD ?
Quiz 1:
 LRT: likelihood ratio test
 The additive (a) and dominance (d) genetic model L0
LRT = −2 ln( )
 P1 (A1A1): m+a; F1 (A1A2): m+d; P2 (A2A2): m-a LA
 LOD : (Likelihood of Odd)
 Code of the three genotypes at a locus: 2 (P1), 0

(P2), 1 (F1)
LA
Genotype P2: A2A2 (0) F1: A1A2 (1) P1: A1A1 (2) LOD = log( ) = log( L A ) − log( L0 )
L0
mid-parent m
Genotypic G22 G12 G11 LRT LRT
value LOD = ≈ LRT ≈ 4.61LOD
2 ln(10) 4.61

Additive a Additive a  A LOD-score threshold of 3 corresponds to a single-test p-value of


approximately 0.0001
Dominance d
20
:What is False Positive?
Quiz 2:
(Shaded area is the distribution of LOD)
0.56
0.5

LOD > LOD0,


LOD False positives, True positives,
i.e., accept
score Type I errors no errors
Ha
0.0 LOD < LOD0, False
i.e., accept True negatives, negatives,
no errors Type II errors
H0
0.0 0.14 0.5
Power = 1- P{Type II
error}
Recombination rate
H0: there is no QTL Ha: there is
at a genomic one QTL at the
LOD>3 taken as evidence of linkage.
position on the genomic
trait in interest position 22

Hypothesis Test αvs Type I Error


Significance Levelα
: Missing the true QTL
 True positive
 True negative  Significance level for N independent tests:
1-(1- α)N
 False positive, Type I error
 False negative, Type II error  Bonferroni correction: ≈ α / N
 When multiple tests are not independent,
Bonferroni correction is too conservative
 Type I error rate= P{ reject H0|H0 is true} and reduce detection power
 Type II error rate= P{ accept H0|H0 is false}  Permutation test

23 24
:What Is QTL
Quiz 3:

Detection Power? :What is Power
Quiz 4:
Analysis ?
 The power of a statistical test is the  Quantify the power of a QTL mapping approach
probability that the test will reject a false  Commonly used to compare different mapping
null hypothesis (i.e., it will not make a methods
Type II error).  In most cases it requires the use of Monte
Carlos Simulation
 Power= 1.0 - Type II error (the
 Generating 100+ mapping populations using the
probability of correctly detecting the true same pre-specified inheritance (QTL) model
QTL)  Conduct QTL detection
 Compute the percentage of correctly detecting the
pre-specified QTL
25 26

Barely Grain weight: DH


population
Example
100 20 40
36 90 60 80
33 IM 100 120
ICIM (PIN=0.01) 80
30 140 160
27 70

Power (%)
Under LOD threshold 3.0, 180 200
24 60
LOD score

IM identified 3 QTL, and 220 240


21 50
18 260 280
ICIM identified 9 QTL 40 300 320
15
12 30 340 360
9 20 380 400
6 420 440
3 10
0 460 480
0 500 520
1H
1H
1H
1H
1H
1H
2H
2H
2H
2H
2H
2H
2H
3H
3H
3H
3H
3H
4H
4H
4H
4H
4H
4H
5H
5H
5H
5H
5H
5H
5H
6H
6H
6H
6H
6H
6H
7H
7H
7H
7H
7H
7H

1 2 3 4 5 10 20 30 FDR 540 560


Testing position on the barley genome, step = 1 cM Percentage of phenotypic variation explained (PVE) 580 600

27 28
:How to Improve
Quiz 5:
Power: Population Size

Mapping Precision?
Probability The length of empirical 95% confidence intervals of QTL
PVE (%) 0.9 0.8 0.7 0.6 PVE (%) PS=200 PS=400
MD=5 MD=10 MD=20 MD=40 MD=5 MD=10 MD=20 MD=40
1 >600 540 cM cM cM cM cM cM cM cM
2 600 420 340 280 1 93.30 104.55 103.10 120.03 47.00 61.94 76.83 88.24
2 54.14 62.37 73.66 86.08 37.55 32.38 38.34 47.59
3 430 280 230 200 3 52.65 47.12 50.02 48.18 25.64 21.95 18.78 33.36
4 340 250 190 160 4 38.89 46.14 41.94 56.72 25.01 18.46 22.74 36.57
5 25.99 37.83 44.73 59.74 16.35 16.39 22.54 36.30
5 280 200 160 130
10 10.31 8.35 11.29 46.45 3.72 4.78 7.92 22.74
10 160 120 100 80 20 8.55 10.19 14.78 26.97 4.90 6.04 8.70 15.56
20 100 80 60 50 30 5.33 8.23 11.56 18.62 3.18 4.78 6.62 12.62

30 100 60 50 40 29 30

How to Improve Mapping


Precision?
:What Is PVE?
Quiz 6:
 Increase population size  PVE = Phenotypic variation explained (%)
 Increase heritability  For single QTL
 Increase marker density: does  PVEg = Vg/Vp * 100%
very high density help?  BC,DH and RILs,Vg=a2 (a is additive effect)
 F2,Vg=a2/2+d2/4 (d is dominant effect)
 Better mapping methods?

31 32
:Can We Add PVEs for
Quiz 7: :Can the Overall PVE Be
Quiz 8:
Each of the QTL Together? More Than 100%?
基因型 频率 基因型值
 Z=X+Y, X and Y are random variables 1
AABB 2
(1 − r ) a1+a2
 E(X+Y) = E(X) + E(Y)
AAbb 1
2
r a1-a2
 V(X+Y) = V(X) + V(Y) +2Cov(X, Y)
aaBB 1
2
r -a1+a2
 When all QTL are independent, Cov(X, Y)=0, aabb 1
2
(1 − r ) -a1-a2
PVEs can be added together to measure
 In RIL,locus A, genetic variance is a12 while for locus B
totally how much varinace is accounted for. it is a22. the total genetic variance is a12 +a22+2(1-2r)a1a2
 When QTL are linked, overall PVE does not  Thus, when QTL are linked, total PVE can be higher than
equal to the sum of individual PVEs 100%; But, for unlinked QTL PVE cannot be higher than
100%
33 34

:What is Distorted
Quiz 9: Segregation Distortion in An
Segregation? Actual Rice F2 Population
 Theoretical segregation ratio
RP129
P1 (AA) X P2 (aa) (no selection) 20

significance level (-log10 P)



RM147
RM159
 P1BC1: AA:Aa=1:1 16

RM491
RP178
 P2BC1: Aa:aa=1:1 RM143
12

RM304
 F2: AA:Aa:aa=1:2:1
8

RM44
RM552
 DH, RIL: AA:aa=1:1
4
 Segregation Distortion
0
 Random drift
1
1
1
1
2
2
2
2
3
3
3
3
4
4
5
5
5
5
6
6
6
6
7
7
8
8
9
9
10
10
11
11
11
12
12
 Gamete/ zygote selection Markers on the 12 rice chromosomes

35
When Distorted Markers Are Not When Distorted Markers Are
Linked With QTL Linked With QTL
A
100 No distortion SDM1 SDM2 SDM3
A C 100
SDM4 SDM5 SDM6 SDM7 100 No distortion SDM1 SDM2 SDM3
A,QTL for 80 SDM8 SDM9
80
SDM4
SDM8
SDM5
SDM9
SDM6 SDM7
80

Power (%)

Power (%)
plant height,

Power (%)
60 60 60

40 40 40
population 20 20
20
size=180 0
0
qPH1-1 qPH1-2 qPH3-1 qPH3-2 qPH4 qPH5 qPH6 qPH7 qPH12
0
qHD1 qHD3 qHD4 qHD5 qHD8 qHD11
B 100 D 120
qPH1-1 qPH1-2 qPH3-1 qPH3-2 qPH4 qPH5 qPH6 qPH7 qPH12 FDR
100
80
B,QTL for B 80

Power (%)
Power (%)
100 60
60
heading 80 40
40

days, 60
20 20
Power (%)

0 0

population 40 qPH1-1 qPH1-2 qPH3-1 qPH3-2 qPH4 qPH5 qPH6 qPH7 qPH12 qHD1 qHD3 qHD4 qHD5 qHD8 qHD11

size=180 20
A,QTL for plant height, population size=180 C, QTL for heading days, population size=180
0 B,QTL for plant height, population size=500 D, QTL for heading days, population size=500
qHD1 qHD3 qHD4 qHD5 qHD8 qHD11 FDR

Effect of Segregation Distortion on :How to Handle


Quiz 10:
QTL Mapping Missing Values ?
 If the distorted marker is not closely linked  Missing marker data
with any plant height or heading date QTL, no  Can be imputed using linkage
significant effects were observed on the relationship
detection power.  Missing phenotypic data
 Otherwise, distorted marker may increase or  Replaced by population mean
decrease the QTL detection power.  Delete the genotype
 In large-size populations, say size of 500, the
effect of distorted marker was minor even the
distorted marker was closely linked with QTL.
40
Power analysis of various Levels
Effect of Missing Markers
(Simulated F2 population from a QTL distribution model and population size 500) of Missing Markers
A
No missing 5% missing 100 0 5% 10% 15% 20% 25% 30% C
100
80
80

Power (%)

Power (%)
60
60
40
40
20 20
0 0
qPH1-1 qPH1-2 qPH3-1qPH3-2 qPH4 qPH5 qPH6 qPH7 qPH12 FDR qHD1 qHD3 qHD4 qHD5 qHD8 qHD11 FDR
B 100
D
100
10% missing 15% missing
80

Power (%)
80

Power (%)
60
60
40
40
20 20

0 0
qPH1-1qPH1-2qPH3-1qPH3-2 qPH4 qPH5 qPH6 qPH7 qPH12 FDR qHD1 qHD3 qHD4 qHD5 qHD8 qHD11 FDR
C
A,QTL for plant height, population size=180 C, QTL for heading days, population size=180
41 B,QTL for plant height, population size=500 D, QTL for heading days, population size=500

:What Is the Proper


Quiz 11: Power and FDR for Two Marker Densities
Marker Density? (whole chromosome as interval)
100 20 14
90 40
12

False discovery rate


60
80
 Experience and simulation studies 70
60
80
100
120
10
8

Power
140
50 160 6
1 marker per 10-20 cM
 40
30
180
200
220
4 10 cM
20 2
240
 Increase marker density increases the 10
0
260
280
0

20
80
140
200
260
320
380
440
500
560
300
chance to make QTL ‘isolated’ 1 2 3 4 5 10 20 30
Phenotypic variation explained (PVE) (%)
320
340 Population size

 Higher marker density is only beneficial if 100 20


40
14
90 12

False discovery rate


population size is sufficiently increased as 80
70
60
80
100
10
60 120 8
well
Power

140
50
40
160
180
6 20 cM
4
 Recombinants are needed between dense 30
20
200
220
240
2
10
markers 0
260
280
0

20
80
140
200
260
320
380
440
500
560
300
1 2 3 4 5 10 20 30 320
 Depends on the mapping methods as well 43
Phenotypic variation explained (PVE) (%) 340 Population size
44
Power and FDR for Two Marker Densities
(10cM interval, QTL in the middle)
:What is QTL by E?
Quiz 12:
100 20 90 A. 互作模式1 B. 互作模式2
90 40 80

False discovery rate


60
80 80 70

基因型值

基因型值
70 100 60
60 120 50
Power

140
50 40
40
160
180 30 10 cM
30 200 20
20
220
10 基因型1 基因型1
240
10 260 0 基因型2 基因型2
0 280

20
80
140
200
260
320
380
440
500
560
300
1 2 3 4 5 10 20 30 320
Phenotypic variation explained (PVE) (%) 340 Population size 环境1 环境2 环境1 环境2

100 20
40
90 C. 互作模式3 D. 互作模式4
90 80

False discovery rate


60
80 80 70
70 100 60

基因型值

基因型值
60 120 50
Power

140
50 160 40 20 cM
40 180 30
30 200
220
20
20 240 10 基因型1 基因型1
10 260 0
0 280 基因型2 基因型2
20
80
140
200
260
320
380
440
500
560
300
1 2 3 4 5 10 20 30 320
Phenotypic variation explained (PVE) (%) 340 Population size 45 环境1 环境2 环境1 环境2 46

Chromosome segments showing :Does the Phenotypic Data Need to


Quiz 13:
QTL for ACE Follow Normal Distribution?
Chromosome 1 2 3 3 5 5 7 7 7 8 8 8 9 10 12 12
F2 B1 B2 F 2 :3 R IL
Segment M4 M18 M21 M23 M35 M39 M49 M50 M51 M54 M56 M57 M59 M69 M77 M79 h mg
2
= 0 .9 0 , h pg
2
= 0 .1 0 h mg
2

2
= 0 .0 0 h mg
2
= 0 .9 5 , h pg
2
= 0 .0 5 h mg
2
= 0 .8 9 , h pg
2
= 0 .1 1 h mg
2
= 0 .8 8 , h pg
2
= 0 .1 2
h = 1 .0 0
LODa E1 0.00 0.26 0.01 0.00 1.87 0.00 0.12 0.03 0.03 0.01 0.02 6.06 0.47 0.01 0.17 0.61 pg

E2 0.00 1.05 2.88 4.66 0.95 0.14 0.35 0.30 0.11 0.13 5.84 3.92 2.89 0.22 0.12 0.05
E3
E4
2.06
0.04
4.03
0.04
0.00
0.00
0.20
4.03
0.84
0.05
2.60
0.32
0.05
0.04
9.61
0.02
4.64
0.49
0.04
0.70
0.04
0.05
3.98
15.89
3.36
6.26
4.35
2.57
0.89
0.10
2.17
0.24
A single major gene and h mg
2
F2
= 0 .8 0 , h pg
2
= 0 .1 0 h
h
mg
B1
2

2
= 0 .0 0
= 0 .3 8
h mg
2
B2
= 0 .8 7 , h pg
2
= 0 .0 5 h
h
F 2 :3
mg
2
2
= 0 .8 2
= 0 .1 2
h mg
2
R IL
= 0 .8 4 , h pg
2
= 0 .1 3
pg pg

E5 0.35 0.66 0.41 3.13 0.40 0.00 4.09 12.51 16.53 2.04 3.18 5.56 9.04 0.31 0.42 0.15 many minor genes: the
E6 0.13 0.38 1.06 1.35 0.30 0.22 0.32 0.02 0.18 0.00 0.16 1.84 4.15 0.90 0.29 0.05
E7 0.08 5.07 0.22 0.21 0.04 0.26 1.47 0.33 0.07 0.03 5.52 6.33 4.07 1.66 4.04 0.01 distribution will be h mg
2
F2
= 0 .7 0 , h pg
2
= 0 .1 0 h
h
mg

pg
B1
2

2
= 0 .0 0
= 0 .2 3
h mg
2
B2
= 0 .7 8 , h pg
2
= 0 .0 5 h
h
F 2 :3
mg

pg
2
2
= 0 .7 4
= 0 .1 2
h mg
2
R IL
= 0 .7 8 , h pg
2
= 0 .1 3

E8 0.01 0.00 0.02 0.45 2.16 1.54 0.01 0.02 0.01 0.67 0.17 16.89 10.01 0.01 0.00 0.00
ADDb E1 -0.11 -0.72 -0.12 -0.08 -2.13 0.06 -0.36 -0.19 0.25 -0.13 0.17 4.13 1.27 0.08 -0.35 -1.99 normal.
E2 -0.07 2.21 -2.98 4.35 -2.11 0.78 -0.88 -0.89 -0.72 0.54 4.43 4.51 4.62 -0.72 -0.43 -0.78 h mg
2
F2
= 0 .6 0 , h pg
2
= 0 .1 0 h mg
B1
2
= 0 .0 0 h mg
2
B2
= 0 .6 9 , h pg
2
= 0 .0 5 h
F 2 :3
mg
2
= 0 .6 6 h mg
2
R IL
= 0 .7 2 , h pg
2
= 0 .1 4
2 2
h = 0 .1 7 h = 0 .1 2
E3 3.89 4.69 0.11 -0.84 -2.03 -3.66 0.32 -6.12 5.11 -0.32 0.35 4.62 5.16 3.57 -1.21 5.68 pg pg

E4 -0.24 -0.23 -0.04 2.05 -0.25 -0.61 -0.16 0.12 0.76 -0.67 -0.18 5.84 3.71 1.32 -0.20 -0.91
E5 1.06 1.22 -0.75 2.42 -0.94 0.07 2.30 -5.19 8.28 -1.59 2.20 3.86 6.40 0.60 -0.55 0.98 However, the residual h mg
2
F2
= 0 .5 0 , h pg
2
= 0 .1 0 h mg
B1
2
= 0 .0 0 h mg
2
B2
= 0 .5 9 , h pg
2
= 0 .0 5 h
F 2 :3
mg
2
= 0 .5 7 h mg
2
R IL
= 0 .6 5 , h pg
2
= 0 .1 6
2 2

E6 -1.22 1.73 -2.29 2.88 -1.53 -1.30 -1.11 -0.30 -1.14 0.13 0.86 3.79 -1.87 -0.85 1.04
needs to be normally
h = 0 .1 3 h = 0 .1 3
7.38 pg pg

E7 -0.52 3.74 -0.56 0.61 -0.31 -0.78 -1.37 -0.69 -0.41 -0.19 3.11 4.29 4.01 1.44 -1.91 -0.21

PVEc
E8
E1
-0.09
0.01
0.05
0.48
-0.11
0.02
0.63
0.01
-1.63
4.15
-1.31
0.00
-0.09
0.23
-0.12
0.06
0.09
0.06
-0.62
0.03
-0.33
0.04
5.94
15.60
4.95
1.00
-0.09
0.01
0.03
0.34
0.09
1.24
distributed. h mg
2
F2
= 0 .4 0 , h pg
2
= 0 .1 0 h
h
mg

pg
B1
2

2
= 0 .0 0
= 0 .1 1
h mg
2
B2
= 0 .4 9 , h pg
2
= 0 .0 5 h
h
F 2 :3
mg

pg
2

2
= 0 .4 7
= 0 .1 3
h mg
2
R IL
= 0 .5 7 , h pg
2
= 0 .1 7

E2 0.00 2.24 6.60 11.44 2.04 0.28 0.69 0.59 0.24 0.26 14.54 9.36 6.66 0.46 0.25 0.10
E3 3.61 7.76 0.01 0.33 1.45 4.73 0.07 21.28 9.19 0.07 0.07 7.54 6.35 8.53 1.53 3.91 F2 B1 B2 F 2 :3 R IL
E4
E5
0.04
0.35
0.05
0.68
0.00
0.41
5.62
3.51
0.06
0.41
0.38
0.00
0.05
4.61
0.03
19.86
0.59
31.26
0.86
2.18
0.05
3.57
34.66
6.79
9.49
12.65
3.35
0.31
0.11
0.41
0.29
0.15
There are methods that h mg
2
= 0 .3 0 , h pg
2
= 0 .1 0 h
h
mg

pg
2

2
= 0 .0 0
= 0 .0 9
h
h
mg

pg
2

2
= 0 .3 8
= 0 .0 6
h
h
mg

pg
2

2
= 0 .3 7
= 0 .1 4
h mg
2
= 0 .4 7 , h pg
2
= 0 .1 9

E6
E7
0.46
0.11
1.37
8.28
3.88
0.29
4.99
0.29
1.07
0.06
0.78
0.36
1.07
2.13
0.07
0.46
0.60
0.10
0.02
0.04
0.55
9.25
6.60
10.89
16.93
6.44
3.06
2.33
0.98
6.37
0.17
0.01
can handle non-normal F2
2
B1
2
B2
2
F 2 :3
2
R IL
2
h mg = 0 .2 0 h mg = 0 .0 0 h mg = 0 .2 6 h mg = 0 .2 6 h mg = 0 .3 4

E8 0.01 0.00 0.02 0.52 2.66 1.70 0.02 0.02 0.01 0.73 0.17 35.08 16.53 0.01 0.00 0.00
47 distribution of residual. h pg
2
= 0 .1 0 h pg
2
= 0 .0 8 h pg
2
= 0 .0 6 h pg
2
= 0 .1 5 h pg
2
= 0 .2 1

48
A simulation: A QTL accounted for 80% of :Do Mapping Component
Quiz 14:
the phenotypic variation with additive effect 1.0 Traits Help?
and located at 25 cM
50
40
Effect Trait I Trait II Addition Subtraction Multiplication Division

Frequency
30 Mean 25 20 45 5 500 1.2563
20
A1 1 0 1 1 20 0.0503
10
0
A2 1 0 1 1 20 0.0503
8.5 9 9.5 10 10.5
Phenotypic value
11 11.5 12 A3 0 1 1 -1 25 -0.0631
100 1.2
A4 0 1 1 -1 25 -0.0631
80 1 A12 0 0 0 0 0 0

Estimated effect
LOD score

60 ICIM 0.8 ICIM A13 0 0 0 0 1 -0.0025


IM 0.6 IM
40
0.4 A14 0 0 0 0 1 -0.0025
20 0.2 A23 0 0 0 0 1 -0.0025
0 0
A24 0 0 0 0 1 -0.0025
100
105
110
115
120
125
130
135
140
145
150
155
160
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
0
5

-0.2

100
105
110
115
120
125
130
135
140
145
150
155
160
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
0
5
1D-scanning on one chromosome, step=1cM 1D-scanning on one chromosome, step=1cM A34 0 0 0 0 0 0.0063
A123 0 0 0 0 0 0
A124 0 0 0 0 0 0
A134 0 0 0 0 0 0.0003
A234 0 0 0 0 0 0.0003
49 A1234 0 0 0 0 0 0 50

Quiz 15: Is Selective Genotyping


Using Composite Trait Reduces
(SG) Useful?
Detection Power and Increase FDR
QTL Trait I Trait II Addition Subtraction Multiplication Division
Model I Power (%) Q1 95.10 69.60 69.30 55.20 50.50
Q2 94.80 69.80 70.40 54.10 50.90
Q3 92.50 67.20 65.30 76.90 75.20
Q4 94.50 68.40 65.40 77.80 75.20
FDR (%) 21.63 22.98 27.42 28.05 28.07 29.68
Model II Power (%) Q1 95.40 67.40 65.60 54.80 49.90
Q2 92.90 62.40 66.00 50.00 49.90
Q3 93.70 69.90 67.00 79.20 74.90
Q4 91.90 62.40 64.90 73.50 72.90
FDR (%) 21.35 22.18 28.76 28.59 28.07 28.89
Model III Power (%) Q1 95.20 66.60 52.40 53.60 37.70 Low High
Q2 95.00 69.20 51.60 54.70 36.40 tail tail
Q3 92.90 63.40 47.80 69.70 56.20 pH − pL
t=
Q4 92.60 61.50 49.90 72.60 58.00
pH (1− pH ) pL (1− pL )
FDR (%) 19.78 23.44 28.83 27.71 29.74 30.18 +
51
2NH 2NL 52
Comparison of SG with IM and ICIM
(PVE= 5%, MD= 5cM and both tails have a selected proportion
of 10%)

SGM has higher detection power than the


conventional IM but lower detection power
than ICIM
SG may still be useful!