epistasis or additive effect

[PAWANDEEP Singh]

Dear Dr. Karl Broman,

I have used the scantwo() function on the RIL population and found a significant hit. In scanone(), one of the QTLs was present, and in scantwo(), there was clear evidence for the second QTL on the same chromosome (based on lodav1 and lodfv1). There was no evidence for interaction, as lodint was not significant, and the effect plot was parallel.  

In this case, the AA:AA and BB:BB genotypes had almost similar phenotypes. BB:AA had a very low trait value, and AA:BB showed the maximum trait value. 

Can we call this case additive with epistasis? 

[Karl Broman]

Statistical geneticists generally use the term “epistasis” to mean any departure from additivity. So I wouldn’t say “additive with epistasis”. I’d either say the loci were additive, or that they showed epistasis. Though actually most of the time we can’t really tell, and so we just say “there is insufficient evidence to conclude that there is epistasis” or “the data are consistent with additivity”.

With your particular data, the question of whether there is epistasis or not depends on the single individual in the BB/AA group (BB at S2_1741558 and AA at S2_2914686). With just one individual in that group, you can’t say anything about additivity vs epistasis in this case: the data are compatible with each.

karl

[PAWANDEEP Singh]

Thank you for the explanation. 

I have a theoretical doubt: when the book talks about the interaction between two loci, does that mean epistasis? Can we see epistasis without interaction (non-significant LODint)? 

In the below example, can we conclude whether it is additivity or epistasis with a very low LODint? (In scanone, S3 (AA) and S10 (BB) had opposite effects, which probably suggests them to be additive)

Traits	chr1	chr2	pos1f	pos2f	lod.full	lod.fv1	lod.int	PVE.full	pos1a	pos2a	lod.add	lod.av1	PVE.add
SADR_1	3	10	97.52	62.23	11.55	3.9	0.02	25.44	97.52	62.23	11.53	3.88	25.4

Thanks and regards,

Pawandeep Singh Kohli

--
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 on the web visit https://groups.google.com/d/msgid/rqtl-disc/6c43b25a-4b66-4b60-8951-1f94e898563fn%40googlegroups.com.

[Karl Broman]

I use the term “epistasis” as the same as “interaction”.

As is typical in hypothesis testing, we can never really conclude that there is no epistasis. We can just say that the data are consistent with additivity. We don’t conclude that the null hypothesis (additivity) is true, but just that we have insufficient evidence to reject the null hypothesis.

karl

[PAWANDEEP Singh]

Thanks for the reply

To view this discussion on the web visit https://groups.google.com/d/msgid/rqtl-disc/0a757356-bec0-456a-916d-d4e29ffbae23n%40googlegroups.com.