Odd Results with StepwiseQTL / Renumbering markers

[sean.t...@gmail.com]

Hello,

I'm getting an odd (maybe) results for QTL on chr 10 after running the following code:

--------

load("AllTests.RData", verbose=T)
load("ds_t1.t2perms.Rdata")
pht<-"ds_t1.t2"

x<-calc.genoprob(x, step=0.5, error.prob=0.05, map.function="kosambi")

step5<-stepwiseqtl(x, pheno.col=pht, method="hk", model="normal", max.qtl=5,
penalties=pens,additive.only=FALSE, refine.locations=TRUE, keeptrace=TRUE, keeplodprofile=TRUE)

summary(fitqtl(x, qtl=step5, formula=formula(step5), pheno.col=pht, method="hk", model="normal", get.ests=T, dropone=T))

---------

fitqtl summary

Method: Haley-Knott regression
Model:  normal phenotype
Number of observations : 152

Full model result
----------------------------------  
Model formula: y ~ Q1 + Q2 + Q3

       df       SS         MS      LOD     %var Pvalue(Chi2) Pvalue(F)
Model   3 3384.068 1128.02260 19.41583 44.46988            0         0
Error 148 4225.730   28.55223                                        
Total 151 7609.798                                                    


Drop one QTL at a time ANOVA table:
----------------------------------  
        df Type III SS    LOD  %var F value Pvalue(Chi2) Pvalue(F)    
8@82.5   1        1408  9.490 18.50   49.30            0  7.38e-11 ***
10@60.1  1        1399  9.439 18.38   48.99            0  8.31e-11 ***
10@60.5  1        1608 10.642 21.13   56.31            0  5.33e-12 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1


Estimated effects:
-----------------
               est       SE      t
Intercept   9.3541   0.4416 21.183
8@82.5      3.0682   0.4370  7.022
10@60.1   -18.8587   2.6943 -7.000
10@60.5    20.2579   2.6997  7.504

-------

for(i in 1:5)
{
print(lodint(step5,qtl.index=i,1.5))
print(lodint(step5,qtl.index=i,1.5, expandtomarkers=T))
}

--------

  chr  pos      lod
c8.loc74.5   8 74.5 7.748817
c8.loc82.5   8 82.5 9.490357
c8.loc84     8 84.0 7.938556
            chr      pos      lod
S8_8946668    8 74.29816 7.448884
c8.loc82.5    8 82.50000 9.490357
S8_25400322   8 84.60799 7.315015
            chr      pos       lod
S10_4607462  10 59.78591 4.8316631
S10_4623022  10 60.08551 9.4386548
S10_4650857  10 60.38906 0.4902432
            chr      pos       lod
S10_4607462  10 59.78591 4.8316631
S10_4623022  10 60.08551 9.4386548
S10_4650857  10 60.38906 0.4902432
  chr      pos       lod
1  10 60.08551  0.000000
8  10 60.50000 10.641652
9  10 61.00000  6.148165
  chr      pos       lod
1  10 60.08551  0.000000
8  10 60.50000 10.641652
9  10 61.00000  6.148165

----------

When I run the following code to check the LOD of individual markers:

---------

plotLodProfile(step5)
l<-attr(step5, "lodprofile")
q1<-l[[1]]
q2<-l[[2]]
q3<-l[[3]]
out<-rbind(q1,q2,q3)
write.csv(out, "swq_Test1.2_LOD.csv")

------------

I see this in .csv file where marker names switch to numbers and my LOD scores at certain positions seemingly have two values (ex. 60.890608 is both LOD 0.49 and 10.03):

Can anyone explain what I'm looking at here? The left most column is my marker, then the chromosome, then position, then LOD score. My troubleshooting hasn't helped. To me it looks like markers "1-7" are just duplicated markers "4623022-4651350" but with different LOD scores. The numbered markers continue on for the rest of the file as well, going all the way up to 444. This is my only phenotype that has produced these results.

Thanks,

Sean

[Karl Broman]

I can't explain the screenshot of the LOD profile output. When you have linked QTL, the LOD profiles for the two QTL will both cover the interval between them, so you'll have a stretch of the chromosome repeated.

When you have a very strong QTL, haley-knott regression can give apparent evidence for two tightly-linked QTL. You might try the multiple imputation method.

karl