Physical vs Genetic Maps: Can Physical Map Be Used?

26 views
Skip to first unread message

Chris Duke

unread,
Jun 6, 2021, 12:01:48 PMJun 6
to R/qtl discussion
Hi all,

I'm using R/QTL to search for SNPs associated with several different phenotypes in double-haploid rainbow trout.

I've worked my way through R/QTL with general success, but have run into an issue when it comes to the genetic map. I have roughly 15,000 screened SNPs from an original set of about 100,000 RAD-seq Illumina SNP data across the 29 chromosomes for my organism, each of which was mapped to a high-density physical location (in BP) based on previous genomic work for the organism (example files attached email). 

R/QTL sends up a warning when I read in the data and clearly knows that my locations are in BP rather than CM, but my solution of omitting the location row on my datasheet during the data read-in (and thus hoping that the program would assume location based on the order of the markers) results in a "Does Not Converge" warning (see below). 

I wanted to ask if you might have a good suggestion for how to proceed. I found a number of FAQs, etc that suggested that using physical maps causes problems, but am not sure how to proceed. I'm not a geneticist by trade, but added a genetic component to my third dissertation chapter, for defense in late July. The FAQ for R/QTL seems to suggest that I need to build a linkage map De Novo, is that necessary for a basic QTL scan? 

Any advice you might be able to offer would be enormously appreciated. I've attached the code that I'm using for the data read-in, example datasets, and the warnings produced for reference. 

Thanks and good luck to everyone,
-Chris Duke


*******************Code & Errors​​​​​​​*****************

QTL1.Length10C <- read.cross("csv", "","D:/FilteredSNP-Length10C-RQTL.csv", genotypes=c("A", "B"), crosstype="dh", estimate.map=TRUE)

 --Read the following data:
87  individuals
14983  markers
2  phenotypes
 --Estimating genetic map
 --Cross type: dh
Warning message:
In summary.cross(QTL1.Length10C) :
  Some chromosomes > 1000 cM in length; there may be a problem with the genetic map.
  (Perhaps it is in basepairs?)

QTL1.Length10C <- read.cross("csv", "","D:/FilteredSNP-Length10C-RQTL-NOBP.csv", genotypes=c("A", "B"), crosstype="dh", estimate.map=TRUE)

 --Read the following data:
87  individuals
14983  markers
2  phenotypes
 --Estimating genetic map
 --Cross type: dh
Warning messages:
1: In est.map(cross, error.prob = error.prob, map.function = map.function) :
  Didn't converge!
2: In summary.cross(cross) :
  Some chromosomes > 1000 cM in length; there may be a problem with the genetic map.
  (Perhaps it is in basepairs?)

Karl Broman

unread,
Jun 6, 2021, 12:03:37 PMJun 6
to rqtl...@googlegroups.com
R/qtl is going to interpret the positions as cM.
As a start, I'd use Mbp rather than bp...divide everything by 10^6.
But if you know the approximate length of the genome in bp and cM, you can assume a constant recombination rate and multiply by total cM / total bp.

I would use estimate.map=FALSE because it is likely time consuming with so many markers.

karl


--
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/62055992-5b45-45a9-98ae-c4cb74940f91n%40googlegroups.com.

Chris Duke

unread,
Jun 6, 2021, 2:16:52 PMJun 6
to R/qtl discussion
Thank you, Dr. Broman, that is extremely helpful. I signed up for the Google group, and resubmitted my question for the group.

Just to confirm that I got my conversion right, if I'm converting a physical position of 339969 bp to cM, it would look like: 

339969 bp/10^6= 0.339969 Mbp

Most estimates of rainbow trout genomes put them at around 2.4 x 10^9 bp and 2927.1 cM in length, so: 

(0.339969 Mbp)*((2927.1 cM)/(2.4 x 10^3 Mbp)) = 0.414634692 cM

Does that look right to you? 

Thank you so much for your help!

-Chris

Karl Broman

unread,
Jun 6, 2021, 2:18:02 PMJun 6
to R/qtl discussion

Yes, that looks right.

karl
Reply all
Reply to author
Forward
0 new messages