Interpreting p-values and LOD scores from fitqtl output.

[Meiling Wong]

Dear Karl,

I am not sure which LOD threshold to use for determining the number of QTL in my model. Here are the results of my analysis:

11. One QTL on chromosome 4A at position 26 was shown in Scanone , with a 5% LOD threshold of 3.39

22. After including the QTL on chr 4A (26) as a covariate in the composite interval mapping, a second QTL was found on the same chromosome at position 286. The 5% LOD threshold for CIM was 3.46.

33. Scantwo: which showed additive effects for the QTL:

 

The "main effect", "heavy" and "light" interaction penalties:

4. The output of stepwiseqtl() suggested 4 QTL, and the fitqtl summary showed that all four QTL are significant (p < 0.05).

Based on the summary from fitqtl, the LOD values for the QTL on Chr 1A and Chr 6A are slightly below the 5% threshold from Scanone (3.39) and CIM (3.46), although the associated p-values are below 0.05. Should I include these two QTL in the final model? What are your thoughts?

Thank you so much and Happy New Year!

Best regards,

Meiling

[Karl Broman]

I would ignore the p-values from fitqtl. They are pointwise p-values and don’t account for the genome scan.

The main effect penalty should really be the same as the scanone threshold, so I would look at why they are so different. Was one based on method=“em” and the other on method=“hk”?

Whether to include the 1A and 6A loci in your model depends on what you’re trying to do, but personally I wouldn’t ignore them. It seems like there’s reasonable but not conclusive evidence for those QTL.

karl

[Meiling Wong]

That was very helpful! 

I realized the difference in LOD values between Scanone and Scantwo was due to the different number of permutations I used (1000 in Scanone vs. 5 in Scantwo). Thank you for pointing it out.

Regarding detecting interactions between QTLs (using the addint function), do you also just consider the LOD scores and ignore the p-values? Thank you again for your insights!

Meiling

--
You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rqtl-disc+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/rqtl-disc/0fa053d4-fc1e-4241-b050-acf2bc4a1c1fn%40googlegroups.com.

[Karl Broman]

Yes, I focus on the LOD scores.

karl

[Meiling Wong]

Thank you so much Karl!

To view this discussion visit https://groups.google.com/d/msgid/rqtl-disc/fd4c8f61-9ac2-4729-a77b-743fecc251e3n%40googlegroups.com.