Question: how to identify which phenotype a variant/kmer is associated with?

[reedus-123]

This isn't an issue but rather a query (apologies if it's in the wrong place).

Having followed the GWAS pipeline, once significant variants/kmers are identified how can one go about identifying which phenotype it's positively associated with?

Thanks in advance, and thanks for creating this software.

[mgalardini]

Hi, could you maybe clarify your question a bit more? The phenotype variants are associated to is the one you provide through the --phenotypes and --phenotype-column command line arguments. The latter is only required if your phenotype is not encoded in the last column. Hope this helps.

[reedus-123]

Thank you for your explanation. What I would like to know is if you have two phenotypes and you find that there is a significant association between variant and phenotype, is there a way to find out which of the two phenotypes the variant is positively associated with? (i.e., found more frequently)

So for instance, if I was comparing two sets of samples (i.e from cattle and from soil) and I found that there was a number of significant variants, how would I know if the presence of these variants is significantly associated with cattle or soil? Also, does the software support the comparison between 3 groups? (i.e., cattle, soil and water)

[mgalardini]

If you want to do associations against a discrete phenotype with multiple classes (> 2), than I'd suggest using something called dummy encoding, or more specifically one-hot encoding. This way you will have 3 phenotypes in your last example (e.g. cattle vs. the rest, soil vs. the rest, and water vs. the rest), and you can run three separate associations. Does that make sense?

[reedus-123]

Yes, it does. So in the phenotypes.pheno file, the structure would be sample phenotype a.fasta 1 b.fasta 2 c.fasta 3

where 1, 2 and 3 are cattle, soil and water respectively.

But my main question is - once I've run the pipeline and concluded that there is a significant association between a variant and phenotype, what steps can I take to figure out which of the three phenotypes the variant is associated with? i.e., is it associated with cattle, soil or water?

Thanks again for answering my questions, I really appreciate it.

[mgalardini]

Hi,

what I meant is something like this:

sample cattle soil water 
a.fasta 1 0 0
b.fasta 0 1 0
c.fasta 0 0 1

[reedus-123]

Oh, perfect, thank you. So if I got this right, the variants identified as significant would be associated with the samples marked as a '1' and not a '0' in their respective column? Is that correct?

[mgalardini]

Yes, that is correct

On Sat, Nov 14, 2020, 23:10 reedus-123 notifications@github.com wrote:

Oh, perfect, thank you. So if I got this right, the variants identified as significant would be associated with the samples marked as a '1' and not a '0' in their respective column? Is that correct?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/mgalardini/pyseer/issues/127#issuecomment-727272084, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAISWXZLVL6X4BSMBHQ2VGDSP353TANCNFSM4TQBQBHA .

[reedus-123]

Sorry about all the questions, thank you for all your help!

[iaposto]

Hello,

I would like to ask a follow-up question; After running a GWAS (--lmm mode, --pres), I see that a lot of significant genes are homogenously distributed among the 0 and 1 phenotypes or even skewed towards 0. This is difficult to interpret given that the significant genes should be associated with the 1 phenotype. Does it makes sense to keep only those with β>0?

Thanks a lot!

[johnlees]

This depends on your phenotype, but generally I would say you should keep them all!

[iaposto]

Thanks for the advice! The phenotype is commensal (0) pathogenic (1).

---

From: John Lees @.> Sent: Tuesday, May 24, 2022 6:56:11 PM To: mgalardini/pyseer @.> Cc: iaposto @.>; Comment @.> Subject: Re: [mgalardini/pyseer] Question: how to identify which phenotype a variant/kmer is associated with? (#127)

This depends on your phenotype, but generally I would say you should keep them all!

— Reply to this email directly, view it on GitHubhttps://github.com/mgalardini/pyseer/issues/127#issuecomment-1136107101, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AHDKLGVFNZHE55WLRVNEXKDVLT3ZXANCNFSM4TQBQBHA. You are receiving this because you commented.Message ID: @.***>