[R] minfi ERROR; return code from pthread

minfi is an R package to analyze DNA methylation arrays, such Illumina 450K and EPIC arrays.

One main function in the minfi package is estimateCellCounts; it can be used to estimate immuno cell fractions from a sample based on DNA methylation values. However, recently when I tried to run the function, I met the following problem during the normalization step:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10


Loading required package: FlowSorted.Blood.450k
Loading required package: IlluminaHumanMethylation450kmanifest
Loading required package: IlluminaHumanMethylationEPICmanifest
[preprocessFunnorm] Background and dye bias correction with noob
Loading required package: IlluminaHumanMethylation450kanno.ilmn12.hg19
[preprocessFunnorm] Mapping to genome
[preprocessFunnorm] Quantile extraction
[preprocessFunnorm] Normalization
Error in preprocessCore::normalize.quantiles.use.target(matrix(crtColumn.> reduced),  :
   ERROR; return code from pthread_create() is 22

I used the parameter ‘processMethod=“preprocessFunnorm”’ in the above run, but using other parameter values led to the same errors.

Causes

After hours of research, I found that this error was caused by different versions of R. Simply, I used the Anaconda R when running the above command, which led to errors. When I swithed to official R installed by following the guide , the program ran smoothly.

A deeper reasearch revealed that the Anaconda R uses an in-built blas from the library libRblas, other than standard/system blas library (such as mkl blas or openblas), even these system libraries are installed already. One can run sessionInfo() and find the used blas library at the section “Matrix products: default”. This libRblas blas seems incompatible with the functions in the minfi package (particularly those using multiple threads).

The R from official R website uses the system blas library (possibly also lapack library).

Solutions

Note, the solution is for Ubuntu Bionic Beaver (18.04;LTS). You need change some values accordingly if using a different system.

So here is the (solution)[https://cloud.r-project.org/bin/linux/ubuntu/README.html], install the official R. The steps are as follows:

put the following lines into /etc/apt/sources.list or a text file in /etc/apt/source.list.d/.

 deb https://cloud.r-project.org/bin/linux/ubuntu bionic-cran35/
 deb http://us-east-1.ec2.archive.ubuntu.com/ubuntu/ bionic-backports main restricted universe

You need change these values to match your system.

run the following commands

 sudo apt-get update
 sudo apt-get install r-base r-base-dev

Finally install minfi by following the bioconductor page. During this step, you may find errors such as the package ‘httr’ isn’t available. If this happens, try to install this manually.

Solutions 2

Updated on 08/19/2022

Another way to fix the issue is to install the right version of BLAS and LAPACK in conda. Here we change the BLAS implementation from openblas to netlib using the following command and it worked for minfi:

1

conda install "libblas=*=*netlib" -c conda-forge

One can find more details about different versions of BLAS implementations at here.

Methods not working

installing bioconductor-minfi package using conda.

Bonus

This page describes more on different versions of BLAS, and how to setup default one with update-alternatives command. Please have a look.

Last modified on 2020-04-10