[Rqtl-disc] Estimated QTL effects with CIM in RIL
384 views
Skip to first unread message
Audrey
Apr 30, 2010, 11:00:31 AM4/30/10
to R/qtl discussion
Hi,
Can you please help me to understanding the following problems?
I used CIM in a RIL family (“risib”). I found 1 QTL on chromosome 3,
position 40. After a scantwo analysis I saw epistatic interactions
with chromosome 2.
I used the function “addqtl” to find one additionnal qtl on the 2nd
chromosome.
First you can maybe tell me if that steps for detection seem correct?
I then used fitqtl to get qtl estimated effects. So if I understand
correctly the epistatic QTL (chr=2, pos=2) is significant, and shows a
negative additive effect of a=-1.18. I don’t understand why it was
not detectable with either scanone or CIM method if his additive
effect is so strong?
The second QTL (chr=3, pos=40) is significant and has a additive
effect of a=1.1
The interaction between both QTL is also significant and has an effect
of a=4.0.
My trait is longevity, so locus of parent 1 on 2-2 decrease the
phenotype by 2.36 days, whereas locus on 3.40 increase longevity by
2.18 days? So most of the phenotypic difference detected here is
caused by epistatic interaction of both QTL (enhance phenotype by 8
days).
The average phenotypic values for the four possible genotypes at the
closest marker are:
qtl2-2 / qtl3-40
AA AA =32.62
AA BB =29.25
BB BB=32.82
BB AA=27.13
If I try to compute roughly the corresponding genetic effects, I find
a = ( (BB AA + BB BB)/2 - (AA AA +AA BB)/2)/2 = -0.46 for locus 1, and
a = +0.58 for locus 2, which do not correspond to the values obtained
in "summary(lod)".
So is my interpretation of estimated effects correct?
Thanks a lot.
> summary(cross)
RI strains via sib matings
No. individuals: 145
No. phenotypes: 1
Percent phenotyped: 100
No. chromosomes: 3
Autosomes: 1 2 3
Total markers: 22
No. markers: 7 6 9
Percent genotyped: 75.7
Genotypes (%): AA:50.2 BB:49.8
>out1 <- addqtl(cross, qtl=qtl, formula=y~Q1, method="hk")
> max(out1)
chr pos lod
c2.loc2 2 2 0.694
> summary(qtl)
QTL object containing genotype probabilities.
name chr pos n.gen
Q1 2@2.0 2 2 2
Q2 3@40.0 3 40 2
>lod <- fitqtl(cross, pheno.col=1, qtl, formula=y~Q1*Q2, method="hk", get.ests=TRUE)
> summary(lod)
fitqtl summary
Method: Haley-Knott regression
Number of observations : 145
Full model result
----------------------------------
Model formula: y ~ Q1 + Q2 + Q1:Q2
df SS MS LOD %var Pvalue(Chi2)
Pvalue(F)
Model 3 686.1743 228.72478 4.675660 13.79983 8.161044e-05
0.0001047072
Error 141 4286.1654
30.39834
Total 144
4972.3398
Drop one QTL at a time ANOVA table:
----------------------------------
df Type III SS LOD %var F value Pvalue(Chi2)
Pvalue(F)
2@2.0 2 225.638 1.615 4.538 3.711 0.024
0.026864 *
3@40.0 2 586.687 4.039 11.799 9.650 9.14e-05
0.000118 ***
2@2.0:3@40.0 1 129.189 0.935 2.598 4.250 0.038
0.041090 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Estimated effects:
-----------------
est SE t
Intercept 30.8884 0.6333 48.776
2@2.0 -2.3577 1.6831 -1.401
3@40.0 2.1862 1.4545 1.503
2@2.0:3@40.0 8.0077 3.8745 2.067
--
You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
To post to this group, send email to rqtl...@googlegroups.com.
To unsubscribe from this group, send email to rqtl-disc+...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/rqtl-disc?hl=en.
Can you please help me to understanding the following problems?
I used CIM in a RIL family (“risib”). I found 1 QTL on chromosome 3,
position 40. After a scantwo analysis I saw epistatic interactions
with chromosome 2.
I used the function “addqtl” to find one additionnal qtl on the 2nd
chromosome.
First you can maybe tell me if that steps for detection seem correct?
I then used fitqtl to get qtl estimated effects. So if I understand
correctly the epistatic QTL (chr=2, pos=2) is significant, and shows a
negative additive effect of a=-1.18. I don’t understand why it was
not detectable with either scanone or CIM method if his additive
effect is so strong?
The second QTL (chr=3, pos=40) is significant and has a additive
effect of a=1.1
The interaction between both QTL is also significant and has an effect
of a=4.0.
My trait is longevity, so locus of parent 1 on 2-2 decrease the
phenotype by 2.36 days, whereas locus on 3.40 increase longevity by
2.18 days? So most of the phenotypic difference detected here is
caused by epistatic interaction of both QTL (enhance phenotype by 8
days).
The average phenotypic values for the four possible genotypes at the
closest marker are:
qtl2-2 / qtl3-40
AA AA =32.62
AA BB =29.25
BB BB=32.82
BB AA=27.13
If I try to compute roughly the corresponding genetic effects, I find
a = ( (BB AA + BB BB)/2 - (AA AA +AA BB)/2)/2 = -0.46 for locus 1, and
a = +0.58 for locus 2, which do not correspond to the values obtained
in "summary(lod)".
So is my interpretation of estimated effects correct?
Thanks a lot.
> summary(cross)
RI strains via sib matings
No. individuals: 145
No. phenotypes: 1
Percent phenotyped: 100
No. chromosomes: 3
Autosomes: 1 2 3
Total markers: 22
No. markers: 7 6 9
Percent genotyped: 75.7
Genotypes (%): AA:50.2 BB:49.8
>out1 <- addqtl(cross, qtl=qtl, formula=y~Q1, method="hk")
> max(out1)
chr pos lod
c2.loc2 2 2 0.694
> summary(qtl)
QTL object containing genotype probabilities.
name chr pos n.gen
Q1 2@2.0 2 2 2
Q2 3@40.0 3 40 2
>lod <- fitqtl(cross, pheno.col=1, qtl, formula=y~Q1*Q2, method="hk", get.ests=TRUE)
> summary(lod)
fitqtl summary
Method: Haley-Knott regression
Number of observations : 145
Full model result
----------------------------------
Model formula: y ~ Q1 + Q2 + Q1:Q2
df SS MS LOD %var Pvalue(Chi2)
Pvalue(F)
Model 3 686.1743 228.72478 4.675660 13.79983 8.161044e-05
0.0001047072
Error 141 4286.1654
30.39834
Total 144
4972.3398
Drop one QTL at a time ANOVA table:
----------------------------------
df Type III SS LOD %var F value Pvalue(Chi2)
Pvalue(F)
2@2.0 2 225.638 1.615 4.538 3.711 0.024
0.026864 *
3@40.0 2 586.687 4.039 11.799 9.650 9.14e-05
0.000118 ***
2@2.0:3@40.0 1 129.189 0.935 2.598 4.250 0.038
0.041090 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Estimated effects:
-----------------
est SE t
Intercept 30.8884 0.6333 48.776
2@2.0 -2.3577 1.6831 -1.401
3@40.0 2.1862 1.4545 1.503
2@2.0:3@40.0 8.0077 3.8745 2.067
--
You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
To post to this group, send email to rqtl...@googlegroups.com.
To unsubscribe from this group, send email to rqtl-disc+...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/rqtl-disc?hl=en.
Audrey
May 10, 2010, 11:01:42 AM5/10/10
to R/qtl discussion
Hi all,
Sorry I think I wasn't really clear, can you confirm me to what the
estimated effects are corresponding to?
As I understood the "est" for each QTL correspond to 2*a ?
For the interaction does it also correspond to twice the additive
effect? or 4 times the additive effect?
Are the estimated effects reliable in RIL design?
Thanks a lot in advance for your help.
Audrey
> Q1 2...@2.0 2 2 2
> Q2 3...@40.0 3 40 2
> 0.026864 *
> 3...@40.0 2 586.687 4.039 11.799 9.650 9.14e-05
> 0.000118 ***
> 2...@2.0:3...@40.0 1 129.189 0.935 2.598 4.250 0.038
> 3...@40.0 2.1862 1.4545 1.503
> 2...@2.0:3...@40.0 8.0077 3.8745 2.067
Sorry I think I wasn't really clear, can you confirm me to what the
estimated effects are corresponding to?
As I understood the "est" for each QTL correspond to 2*a ?
For the interaction does it also correspond to twice the additive
effect? or 4 times the additive effect?
Are the estimated effects reliable in RIL design?
Thanks a lot in advance for your help.
Audrey
> Q2 3...@40.0 3 40 2
>
> >lod <- fitqtl(cross, pheno.col=1, qtl, formula=y~Q1*Q2, method="hk", get.ests=TRUE)
> > summary(lod)
>
> fitqtl summary
>
> Method: Haley-Knott regression
> Number of observations : 145
> Full model result
> ----------------------------------
> Model formula: y ~ Q1 + Q2 + Q1:Q2
>
> df SS MS LOD %var Pvalue(Chi2)
> Pvalue(F)
> Model 3 686.1743 228.72478 4.675660 13.79983 8.161044e-05
> 0.0001047072
> Error 141 4286.1654
> 30.39834
> Total 144
> 4972.3398
>
> Drop one QTL at a time ANOVA table:
> ----------------------------------
> df Type III SS LOD %var F value Pvalue(Chi2)
> Pvalue(F)
> 2...@2.0 2 225.638 1.615 4.538 3.711 0.024
> >lod <- fitqtl(cross, pheno.col=1, qtl, formula=y~Q1*Q2, method="hk", get.ests=TRUE)
> > summary(lod)
>
> fitqtl summary
>
> Method: Haley-Knott regression
> Number of observations : 145
> Full model result
> ----------------------------------
> Model formula: y ~ Q1 + Q2 + Q1:Q2
>
> df SS MS LOD %var Pvalue(Chi2)
> Pvalue(F)
> Model 3 686.1743 228.72478 4.675660 13.79983 8.161044e-05
> 0.0001047072
> Error 141 4286.1654
> 30.39834
> Total 144
> 4972.3398
>
> Drop one QTL at a time ANOVA table:
> ----------------------------------
> df Type III SS LOD %var F value Pvalue(Chi2)
> Pvalue(F)
> 0.026864 *
> 3...@40.0 2 586.687 4.039 11.799 9.650 9.14e-05
> 0.000118 ***
> 2...@2.0:3...@40.0 1 129.189 0.935 2.598 4.250 0.038
> 0.041090 *
> ---
> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
> Estimated effects:
> -----------------
> est SE t
> Intercept 30.8884 0.6333 48.776
> 2...@2.0 -2.3577 1.6831 -1.401
> ---
> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
> Estimated effects:
> -----------------
> est SE t
> Intercept 30.8884 0.6333 48.776
> 3...@40.0 2.1862 1.4545 1.503
> 2...@2.0:3...@40.0 8.0077 3.8745 2.067
Karl Broman
May 12, 2010, 11:56:42 PM5/12/10
to rqtl...@googlegroups.com, Audrey
For a backcross, the estimated additive effect for a QTL in fitqtl() is the estimated difference between the average phenotype for the heterozygote and that for the homozygote. Basically we code the two genotypes as +/- 0.5 and then do linear regression.
For RIL, it should be half the difference between the two homozygotes, but it looks like it's currently giving just the difference between the two homozygotes. I'd call that a bug.
The interaction effects are what you get when you multiply the two QTL genotype columns together. You then can see that the 2x problem in the additive effect leads to a 4x bug in the interactive effect. It's supposed to be the difference in the additive effects at one QTL when you consider separately the two genotypes at the other QTL.
karl
For RIL, it should be half the difference between the two homozygotes, but it looks like it's currently giving just the difference between the two homozygotes. I'd call that a bug.
The interaction effects are what you get when you multiply the two QTL genotype columns together. You then can see that the 2x problem in the additive effect leads to a 4x bug in the interactive effect. It's supposed to be the difference in the additive effects at one QTL when you consider separately the two genotypes at the other QTL.
karl
Reply all
Reply to author
Forward
0 new messages