How to analyze genetic data that includes relatives?

[fbbriggs]

Does anyone have suggestions? I have a relative small dataset for a rare phenotype, as a result I would like to maximize the sample size retained (versus randomly retaining only one relative as it standard in traditional GWAS)?

I have heard of REGENIE and GENESIS? Any experience or other suggestions?

[jgottesman]

@hirotaka or @mike any thoughts?

[mike]

Any mixed model that incorperates kinship from GCTA to regenie will do the job.

[hirotaka]

I haven’t conducted analyses that include relatives, but these softwares (https://yanglab.westlake.edu.cn/software/gcta/#Overview) and GENESIS seem capable of such analyses with GRM.

[paularp]

I believe @peixott is currently using SAIGE! Home | SAIGE

[peixott]

@fbbriggs

I have heard of REGENIE and GENESIS? Any experience or other suggestions?

I haven’t conducted analyses that include relatives, but these softwares (https://yanglab.westlake.edu.cn/software/gcta/#Overview) and GENESIS seem capable of such analyses with GRM.

I am using SAIGE and TRACTOR. In a recent benchmark [1] showed that SAIGE is the best in type of Type-I error rate and TRACTOR is the most powerful analysis for admixed (and LARGE-PD is all about admixed people).

SAIGE allows you to include the genetic relationship matrix (GRM) in the model, so you do not remove anyone based on relationship. TRACTOR does not use GRM, so we need to remove related people. To do this I use NAToRA [2] (which I am the creator, so I may be biased) that is better than PLINK and KING.

I hope to help
Thiago Peixoto Leal

Ref:
[1] DOI: 10.1093/bib/bbad437
[2] DOI: 10.1016/j.csbj.2022.04.009