Bayes Credible Interval is zero or negative?

[Lena Flörl]

Hi all, 

for some of my QTLs the Bayes Credible Interval (BCI) is between the same two markers - so the effective interval is zero. And I'm afraid I don't understand what that means. Is that a fake QTL at a single marker or just a very sharply defined one?

Upon calculating the LOD support intervals for the same QTLs that have a BCI of zero, I sometimes get very small ones as well (e.g. 76 bp) or fairly big ones (~100 kpb). 

Thanks! 

Cheers!

Lena 

[Karl Broman]

You apparently have a very sharp peak. I can’t say whether it’s real or fake. You might re-run the analysis after first inserting more pseudomarkers, by running calc.genoprob with a small positive value for step.

karl

[Lena Flörl]

Dear Karl, 

after over a year I have a follow up question regarding these sharply defined intervals! 

My intervals are indeed super sharp (e.g. 2e-4 cM). However, to parse genes I need to translate this interval to a physical location. Repeating the BCI calculation without pseudomarkers therefore always gives me a start and stop interval marker which are switched ie. the stop marker lies before the start marker in the genome - hence the original negative intervals. 

Maybe for reference, this is particularly about a hotspot on a chromosome where I get multiple different significant QTLs which are 200+ kbp apart.  When I calculate the BCI with pseudo markers the interval size in cM varies quite a bit, but overall is rather small. However when I repeate the BCI calculation without pseudo markers I get the same/similar physical markers delineating the intervals for different significant QTLs. 

What's more is that I always get the same stop interval marker - which makes me wonder whether this could have any relevance? 

Now my question is whether there is a rule of thumb with which I could convert a cM distance to bp distance?

I am absolutely aware that these are different units, and the linkage varies strongly by region.

But do you think it’s okay to do a rough estimate, e.g. given that my reference genome has an approximate size of 495 Mb and we have 5386 markers with a median distance of 28.96±22.25 cM, to use e.g. 3000 bp per cM?

Alternatively I thought of just using the flanking markers surrounding a significant QTL to search for genes within - but as the BCI intervals do vary over the different significant QLTs this does not seem reliable either.

I also tried to just switch the start and stop marker delineating the support interval if one happens to lie before the other - however the retrieved intervals and thereby number of genes is just unnecessarily large, particularly since my BCI actually seems so sharply defined. 

Thank you so much! 

I really appreciate any insight! 

Kind regards,

Lena 

[Karl Broman]

Particularly when calculating Bayesian intervals, it is important to include dense pseudomarkers, as in calculating the interval you’re needing to approximate areas with sums, and if you just include the markers themselves those approximate areas will be very crude.

In the case of very sharply defined QTL, it would be good to use step sizes well below 1 cM.

If you want the endpoints of the intervals to be markers, so that they are tagged as physical locations, use expandtomarkers=TRUE. 

The approach I’ve generally used, to get bp locations, is to use interpolation between the genetic and physical maps, for example with the interpPositions() function.

When you’re looking at specific intervals and trying to relate cM distances to bp distances, you definitely don’t want to rely on overall estimates of recombination rate, because the recombination rate can have so much local variation in most organisms.

karl

[Lena Flörl]

Dear Karl,

thanks a lot - that's super helpful! I have used the interpolation function to estimate the location of my pseudo markers and it looks much better. 

Something I still notice is that sometimes my Bayes Credible intervals are substantially shifted. For example, I get an interval of 0.1 cM and the start and stop marker are close to each other, but lie 750 kb upstream of the actual significant QTL. Is there an explanation for this behaviour? 

Thank you and Kind regards,

Lena 

[Karl Broman]

The interval should contain the peak LOD. I can't explain what is going on without seeing the details.

karl