Building a genetic map from RAD-seq using OneMap

[standonn]

Dear all,

I have some RAD-seq data from 2 parents and 97 offsprings. I would like to use this data to cluster my genomic scaffolds into linkage groups. I'm new to handling RAD-seq data and I'm not sure if what I'm doing is correct.

For the moment I have done the following:
- Quality Control of the RAD-seq reads
- Demultiplexed the files
- Mapped the RAD-seq reads to the genomic scaffolds (Bowtie2 + Samtools)
- Ran the ref_map.pl pipeline of Stacks, specifying a "onemap" output from the program "genotypes"
- Followed the tutorial step of OneMap (https://cran.r-project.org/web/packa...ed_version.pdf)

Now here is where things start not working: as I increase the LOD value, I get more linkage groups but I also increase the number of unlinked markers.

For example, with LOD=6, I get:
> groups
This is an object of class 'group'
It was generated from the object "all.mark"

Criteria used to assign markers to groups:
LOD = 6 , Maximum recombination fraction = 0.5

No. markers: 1342
No. groups: 1
No. linked markers: 901
No. unlinked markers: 441

With LOD=20, I get:
> groups
This is an object of class 'group'
It was generated from the object "all.mark"

Criteria used to assign markers to groups:
LOD = 20 , Maximum recombination fraction = 0.5

No. markers: 1342
No. groups: 9
No. linked markers: 225
No. unlinked markers: 1117

Is it normal to lose so many markers? What am I doing wrong?

Also some groups (with LOD=20), only have 2 markers which I guess is insufficient to have a good genetic map.

Any help or insight about building a genetic map using RAD-seq highly appreciated!

Cheers!

[SNPsaurus]

You could write your question to the Stacks Google group (https://groups.google.com/forum/#!forum/stacks-users). Julian is pretty responsive.

I'd say it is normal to increase the number of unlinked markers as you increase the LOD score stringency. I don't know if your numbers look typical though.

[standonn]

Thanks for answering! I'll follow your advice and post my question on the Stacks forum. Thanks for the link!

[fenduoduo]

Hi standonn,
I have the same problem. Do you have a solution now?

[standonn]

Yes, the problem was solved the following way:

First, after demultiplexing the RAD-seq data, I ran the denovo_map of stacks. I specified "F2" as the cross (option -A). Depending on the cross that was performed choose the one corresponding to your data. I think this was my problem. I didn't choose the right cross thinking I had another type of data. Once that confusion was sorted, everything started to work.

After running deno_map.pl I ran the genotypes program the following way:
genotypes -b 1 -P . -r 7 -t F2 -o onemap -c

This allowed me to get the genotypes for each marker and of each progeny sample in the onemap format.

Then I ran onemap to build linkage groups and order the markers:
This is what I got:
> F2 <- read.mapmaker(file="for-onemap-analysis.txt")
--Read the following data:
Type of cross: f2
Number of individuals: 95
Number of markers: 1329
> twopts.f2 <- rf.2pts(F2)

> mark.all <- make.seq(twopts.f2, "all")
> (LGs <- group(mark.all, LOD=20, max.rf=0.5))
This is an object of class 'group'
It was generated from the object "mark.all"

Criteria used to assign markers to groups:
LOD = 20 , Maximum recombination fraction = 0.5

No. markers: 1329
No. groups: 7
No. linked markers: 1294
No. unlinked markers: 35

With this LOD/max.rf I built 7 groups (and I know this is the number of chromosome of my species). These groups contain most of the markers. I then continued the analysis: ordered the markers using the RECORD method, cleaned the map using R/QTL (with R/QTL you can remove duplicate markers and samples, remove markers/samples with a lot of missing data, look at the segregation distortion...).

I hope this helps. Ah, I also re-ran the stacks ref_map.pl pipeline and it gave me the same result (the maps from the denovo_map.pl and from ref_map.pl are extremely similar).

Best,
Sophie