polyRAD

[xibeixingchen]

Hello Lindsay,

I hope this email finds you well.

I am currently conducting research on population genetics in autotetraploid alfalfa. My research species is a highly heterozygous, outcrossing autotetraploid. We have whole-genome resequencing (WGS) data for a natural population, with individual average sequencing depths ranging from 10x to 30x.

Our primary objectives are to perform accurate allele dosage estimation and subsequently conduct Genome-Wide Association Studies (GWAS) based on these dosage calls. As I move forward with this work, I have encountered some specific technical questions and would be very grateful for your expert advice:

1. Application of polyRAD Software: We plan to use polyRAD for allele dosage estimation. I would like to ask, in your experience, at what sequencing depth does polyRAD typically maintain high accuracy (i.e., a high true positive rate) for allele dosage estimation in autotetraploid species? Do you have any relevant experience or recommended depth thresholds for this?

2. Strategies for Low Sequencing Depth (approx. 10x) Samples: A subset of individuals in our population has a sequencing depth of around 10x, and we are concerned this might compromise the accuracy of dosage estimation. We have two initial ideas on how to address this and would greatly appreciate your professional opinion:

   • Idea 1: Imputation using a high-coverage reference panel. This would involve selecting individuals with higher sequencing depth (e.g., >20x or >25x) from our population to create a reference panel, and then imputing genotypes/dosages for the low-depth (approx. 10x) individuals. Is this strategy feasible and effective in highly heterozygous autotetraploids? If so, are there any recommended imputation tools or parameter settings for such a scenario?
   • Idea 2: GATK4 joint calling. During variant discovery and genotyping, would using tools like GATK4 for joint calling across the population provide benefits from shared information, thereby potentially improving allele dosage estimation accuracy for the low-depth individuals?
   • Alternatively, are there other more effective methods or suggestions you might have for improving the quality of allele dosage estimation for these low-depth samples?

3. Comparison of polyRAD vs. Updog: We have also noted the Updog software, which is also used for polyploid genotype/dosage estimation. Considering our species (autotetraploid alfalfa, highly heterozygous, outcrossing) and data characteristics (WGS, 10x-30x depth), which software, polyRAD or Updog, do you think might be more suitable for our needs? Or, what are their respective main advantages and limitations when handling such data?

These questions may be quite specific, and I am very grateful for you taking the time out of your busy schedule to provide guidance. Any advice you can offer would be invaluable to my research.

Please let me know if you require any further details about our data or research design.

Thank you once again for your time and help!

Sincerely,

Zhicheng

[lvclark]

Hi Zhicheng,

If you need high accuracy in your genotype calls, you need hundreds of reads in a tetraploid, and at that point the genotype calling algorithm/software doesn't matter so much. polyRAD and updog are both designed to make the best out of low accuracy situations.

PolyRAD can use population structure and linkage disequilibrium to help impute genotypes like you mention. I don't think that GATK would do that.

In general terms, I think that polyRAD is better at dealing with biological issues (like population structure) and updog is better at dealing with technical issues like library amplification. Which works better depends on the dataset. Last I checked, polyRAD ran about 100x faster however.

I have changed career paths somewhat since creating polyRAD, so I'm not familiar with as many use cases as you would probably like. I hope my input is helpful though!

Best,

Lindsay