Bootstrapping phenotypes to create a null distribution

[Saikat Bandyopadhyay]

I have QTL and eQTL data on a DO population. 

I am conducting a conditional analysis, where I condition the QTL analysis on the allele dosage of eQTL lead SNPs (an additive covariate) to compute the ΔLOD.

I am trying to answer the following question

"How much ΔLOD would we expect simply from sampling variability while conditioning on the specific SNP we care about?"

To answer this, I am considering a bootstrapping-like approach, where I create pseudo-datasets using the following code:

pheno_vec <- as.data.frame(cross_basic$pheno)[, pheno_id]

names(pheno_vec) <- rownames(cross_basic$pheno)

boot_ids <- sample(seq_len(length(pheno_vec)), replace = T)
pheno_boot      <-  pheno_vec[boot_ids]

In the pseudo dataset created by this approach, some of the samples are repeated.  I am not able to use this bootstrap vector to run a QTL scan (Duplicate names in argument 2). Is there a correct way of achieving this? What other approaches could I use?

I thought of doing `names(pheno_boot) <- names(pheno_vec)`, but I am worried, it breaks the genotype-phenotype association of the samples.

[Dan Gatti]

I'm not sure, but would a permutation approach work? That way you wouldn't have duplicate sample labels. 

Or, if you want to use bootstrap, rename your duplicated samples by appending a number or character to make them unique.

Dan

---

From: rqtl2...@googlegroups.com <rqtl2...@googlegroups.com> on behalf of Saikat Bandyopadhyay <saikat....@gmail.com>
Sent: Thursday, September 11, 2025 4:15 PM
To: R/qtl2 discussion <rqtl2...@googlegroups.com>
Subject: [SOCIAL NETWORK] [Rqtl2-disc] Bootstrapping phenotypes to create a null distribution

 

--
You received this message because you are subscribed to the Google Groups "R/qtl2 discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rqtl2-disc+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/rqtl2-disc/98ea67fc-bca6-4e61-9061-91ec1cc3f8c0n%40googlegroups.com.

---

The information in this email, including attachments, may be confidential and is intended solely for the addressee(s). If you believe you received this email by mistake, please notify the sender by return email as soon as possible.

[Saikat Bandyopadhyay]

Thanks, Dan, for the quick response.

I have also considered using a permutation approach, and it seems more sensible to me. Just to confirm, are you saying that a permutation approach will also yield the same answer?

The bootstrap method was suggested to me. However, I have concerns that duplicate samples will dilute the genetic signal in every iteration of the bootstrap, as some samples are repeated. And I wonder if that impacts the interpretability of the quantile range that we get by bootstrapping?

I would be grateful if you could address these. Thanks again!