methods of genome scanning

[nata...@gmail.com]

Hello,

I used the EM method for both normal and binary traits in scanone/ scantwo. But in addqtl/ addpair there are different methods. 

Is it ok to use different methods as long as I mention which methods I used for each step? 

Also, can I use EM for binary traits and "hk" for the normal traits?

Running scantwo with the 2 binary traits is time consuming and it seems like this mehod ("hk") does not fit the binary model. 

what is a good LOD score for addqtl/ addpair? my max LOD of addqtl's output was 1.5

thanks,

Nataly

[Karl Broman]

There is no requirement that you use the same methods in the single-QTL genome scan from scanone() as you use with scantwo() or multiple-QTL mapping methods. But it would be easier to explain if you stuck to the same method, and different methods might need different thresholds. If you get similar results from Haley-Knott regression as the EM algorithm in scanone(), then I'd just stick with Haley-Knott regression, as that's what is implemented in fitqtl() and addqtl().

Note that method="hk" or ="em" has to do with how the model is fit, and this is separate from model="normal" or ="binary". You can use model="normal" or ="binary" in scantwo and addqtl.

Regarding LOD scores for addqtl/addpair, it depends on the size of the genome and the type of cross. I prefer to base the threshold on the results of a permutation. See Broman and Speed (2002) https://doi.org/10.1111/1467-9868.00354 and Manichaikul et al. (2009) https://doi.org/10.1534/genetics.108.094565

karl

[nata...@gmail.com]

thank you for your answer

I understand that it's ok to use different methods for binary and normal traits. the results for normal traits look the same with both methods. 

out.sc1<-scanone(mycross, pheno.col=c(1:11), method="hk", model="normal")

out.sc1.bin<-scanone(mycross, pheno.col=c(12:13), method="em", model="binary")

for the addqtl I have to use perms of the scanone results?

ב-יום חמישי, 25 בפברואר 2021 בשעה 00:08:58 UTC+2, kbr...@gmail.com כתב/ה:

[Karl Broman]

If you’re not considering interactions, then the approach I have recommended is to use a fixed threshold for each additional QTL, the one you get from acanone permutations.

karl 

[nata...@gmail.com]

Hi again and thanks for your reply,

for normal traits I used scantwo with a 2 cM step

for binary traits I used a wider step (10 cM) for scantwo(), should I consider the step when making a qtl object in order to use addpair()?

ב-יום חמישי, 25 בפברואר 2021 בשעה 16:17:54 UTC+2, kbr...@gmail.com כתב/ה: