How to account for relatedness in plink?

1,378 views
Skip to first unread message

teklu.t...@gmail.com

unread,
Mar 4, 2019, 4:04:42 PM3/4/19
to plink2-users

Hi Christopher,


I am using plink for GWAS in plants and I have some questions regarding accounting for relatedness. I used multidimensional scaling implemented in Plink to correct for population structure but the Q-Q plots are not that great because MDS did not account for relatedness. In TASSEL, there is an option to correct for both population structure and kinship but I am not sure how to account for kinship in plink. Previous authors who used plink in human genetics suggested removing individuals from each pair with identity by descent > 0.25 (second degree relatives). But in plants almost all lines have an IBD higher than that and removing lines is not a solution. Is there a way to account for both population structure and kinship in plink? 


I run haplotype based GWAS in plink and the output looks like this:


NSNP NHAP  CHR          BP1          BP2             SNP1             SNP2    HAPLOTYPE        F     BETA     STAT        P
 2    3    1       994893       994908    LcChr1p994893    LcChr1p994908           TC   0.0914  0.00122 1.61e-005    0.997
 2    3    1       994893       994908    LcChr1p994893    LcChr1p994908           CC    0.191   0.0864    0.134    0.715
 2    3    1       994893       994908    LcChr1p994893    LcChr1p994908           CT    0.718  -0.0755    0.118    0.732
 2    3    1      1024189      1024637   LcChr1p1024189   LcChr1p1024637           GT    0.258  -0.0896    0.253    0.615
 2    3    1      1024189      1024637   LcChr1p1024189   LcChr1p1024637           AC    0.237    0.381     3.85   0.0506
 2    3    1      1024189      1024637   LcChr1p1024189   LcChr1p1024637           GC    0.503   -0.243     1.87    0.172

How do I get the proportion of variance (R2) explained by each haplotype window?

Thank you so much in advance for your time and help.

Christopher Chang

unread,
Mar 4, 2019, 4:24:06 PM3/4/19
to plink2-users
You can account for relatedness by using plink 2.0 --make-bed + --king-table-filter <threshold> before the main association analysis, to prune away close relations.  A threshold of 0.177 corresponds to removing 1st-degree relations, and 0.088 also removes 2nd-degree relations.  This is a lot more reliable than anything available in plink 1.x.

Population structure is typically accounted for by using top principal components (computed via plink --pca, or a similar function in another software package) as covariates.  Unfortunately, plink 1.07's haplotype association analysis does not support covariates; but the main allelic association command (--linear/--logistic, or --glm in plink 2.0) does support them.
Message has been deleted

Christopher Chang

unread,
Mar 4, 2019, 6:51:38 PM3/4/19
to plink2-users
You can try linear mixed model association analysis; this isn't built into plink, but there are several other software packages which can perform this on plink-formatted files.

Thank you for the swift reply. In my population all individuals are higher than second degree relatives. So if I remove them based on the threshold of 0.177 then I will discard 90% of the individuals. Is there a way to account for relatedness without removing close relatives?

Regards,
Message has been deleted

Teklu Teklay

unread,
Mar 28, 2019, 5:16:03 PM3/28/19
to plink2-users
Hi Christopher,
I calculate haplotype blocks and wanted to get individual haplotype phases. I used --hap-phase in plink 1.7 but this outputs phased haplotypes for each block separately. Is there a way to get all phased haplotypes (most likely phase for each individual) for all individuals in a single file?
Thanks Again,


On Mon, 4 Mar 2019 at 18:20, <teklu.t...@gmail.com> wrote:
Ok thank you.

--
You received this message because you are subscribed to the Google Groups "plink2-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plink2-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Christopher Chang

unread,
Mar 28, 2019, 5:21:50 PM3/28/19
to plink2-users
plink 1.07's phasing functions are no longer supported.  Use other software for this.


On Thursday, March 28, 2019 at 2:16:03 PM UTC-7, Teklu Teklay wrote:
Hi Christopher,
I calculate haplotype blocks and wanted to get individual haplotype phases. I used --hap-phase in plink 1.7 but this outputs phased haplotypes for each block separately. Is there a way to get all phased haplotypes (most likely phase for each individual) for all individuals in a single file?
Thanks Again,

Ok thank you.

On Monday, 4 March 2019 17:51:38 UTC-6, Christopher Chang wrote:
You can try linear mixed model association analysis; this isn't built into plink, but there are several other software packages which can perform this on plink-formatted files.

Thank you for the swift reply. In my population all individuals are higher than second degree relatives. So if I remove them based on the threshold of 0.177 then I will discard 90% of the individuals. Is there a way to account for relatedness without removing close relatives?

Regards,


On Monday, 4 March 2019 15:24:06 UTC-6, Christopher Chang wrote:
You can account for relatedness by using plink 2.0 --make-bed + --king-table-filter <threshold> before the main association analysis, to prune away close relations.  A threshold of 0.177 corresponds to removing 1st-degree relations, and 0.088 also removes 2nd-degree relations.  This is a lot more reliable than anything available in plink 1.x.

Population structure is typically accounted for by using top principal components (computed via plink --pca, or a similar function in another software package) as covariates.  Unfortunately, plink 1.07's haplotype association analysis does not support covariates; but the main allelic association command (--linear/--logistic, or --glm in plink 2.0) does support them.

Hi Christopher,


I am using plink for GWAS in plants and I have some questions regarding accounting for relatedness. I used multidimensional scaling implemented in Plink to correct for population structure but the Q-Q plots are not that great because MDS did not account for relatedness. In TASSEL, there is an option to correct for both population structure and kinship but I am not sure how to account for kinship in plink. Previous authors who used plink in human genetics suggested removing individuals from each pair with identity by descent > 0.25 (second degree relatives). But in plants almost all lines have an IBD higher than that and removing lines is not a solution. Is there a way to account for both population structure and kinship in plink? 


I run haplotype based GWAS in plink and the output looks like this:


NSNP NHAP  CHR          BP1          BP2             SNP1             SNP2    HAPLOTYPE        F     BETA     STAT        P
 2    3    1       994893       994908    LcChr1p994893    LcChr1p994908           TC   0.0914  0.00122 1.61e-005    0.997
 2    3    1       994893       994908    LcChr1p994893    LcChr1p994908           CC    0.191   0.0864    0.134    0.715
 2    3    1       994893       994908    LcChr1p994893    LcChr1p994908           CT    0.718  -0.0755    0.118    0.732
 2    3    1      1024189      1024637   LcChr1p1024189   LcChr1p1024637           GT    0.258  -0.0896    0.253    0.615
 2    3    1      1024189      1024637   LcChr1p1024189   LcChr1p1024637           AC    0.237    0.381     3.85   0.0506
 2    3    1      1024189      1024637   LcChr1p1024189   LcChr1p1024637           GC    0.503   -0.243     1.87    0.172

How do I get the proportion of variance (R2) explained by each haplotype window?

Thank you so much in advance for your time and help.

--
You received this message because you are subscribed to the Google Groups "plink2-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plink2-users+unsubscribe@googlegroups.com.

Hannah Scheppler

unread,
Jun 20, 2024, 1:57:20 PM6/20/24
to plink2-users
Does --king-table-filter prune away variants or whole samples? 
Does plink2 have an option to utilize the kinship matrix directly in the --glm? So that we can account for relatedness in the model without removing data first?

Thank you,
Hannah

Matthew Maher

unread,
Jun 20, 2024, 9:58:14 PM6/20/24
to Hannah Scheppler, plink2-users
--king-cutoff can be used to prune out related samples prior to e.g. --glm

PLINK does not support mixed-model needed in order to keep and account for sample-relatedness.  For that, check out the tools  REGENIE or SAIGE

--
You received this message because you are subscribed to the Google Groups "plink2-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plink2-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plink2-users/762288cb-26d9-45d6-a306-e180e115fb25n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages