Question: ADMIXTURE haploid mode does not converge
gravatar for samuelandjw
22 months ago by
samuelandjw20 wrote:

I'm trying to reproduce the results (Table 1) described in

Shringarpure et al. - 2016 - Efficient analysis of large datasets and sex bias

The results involved calculation of global admixture proportions on the X chromosomes of both male and female samples, using the newly introduced "haploid" mode. I downloaded the chrX genotypes in PLINK1 format from here. I filtered the genotypes as described in the paper, by removing SNPs with MAF<0.05 and LD thinning pairs of SNPs with r^2<0.1. After that, I ran ADMIXTURE haploid mode using option --haploid=“male:23”. However, the run had not converged after >100 iterations.

This is the script I ran ADMIXTURE:

plink --bfile 1kg_phase1_chrX
--memory 2000
--maf 0.05
--make-bed --out chrX.maf_0.05

plink --bfile chrX.maf_0.05 --indep-pairwise 50 5 0.1

plink --bfile chrX.maf_0.05 --extract --make-bed --out chrX.filtered admixture --haploid="male:23" -s 1 chrX.filtered.bed 2 -j 12

Looking at the log files, it seems that ADMIXTURE had a hard time in finding a local maximum:

276 (QN/Block) Elapsed: 1.811 Loglikelihood: -8.52897e+06 (delta): 1.42774
277 (QN/Block) Elapsed: 1.602 Loglikelihood: -8.52881e+06 (delta): 156.512 278 (QN/Block) Elapsed: 1.685 Loglikelihood: -8.52891e+06 (delta): -99.4092 279 (QN/Block) Elapsed: 1.618 Loglikelihood: -8.52884e+06 (delta): 68.4685 280 (QN/Block) Elapsed: 1.653 Loglikelihood: -8.52883e+06 (delta): 12.983 281 (QN/Block) Elapsed: 1.697 Loglikelihood: -8.52894e+06 (delta): -110.379

What did I do wrong?

X admixture population • 617 views
ADD COMMENTlink modified 22 months ago • written 22 months ago by samuelandjw20
gravatar for samuelandjw
22 months ago by
samuelandjw20 wrote:

After communications with experts in the filed, it seems that there is bug with ADMIXTURE when both --haploid and -j (multi-threading flag) are used at the same time. Using --haploid without using -j solves the problem.

ADD COMMENTlink written 22 months ago by samuelandjw20
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1074 users visited in the last hour