Hi all,

I am trying to use PEER in R (on conda) to identify hidden confounders in the raw counts of my RNA-seq experiment. I am following their tutorial ( https://github.com/PMBio/peer/wiki/Tutorial ), but without much luck, and there's not much help available unfortunately. Anyone used this tool before and could advise?

I have installed R-PEER on Conda, but am having issues with following with the PEER pipeline, to identify 10 hidden confounders in my expression data. I was wondering if anyone could advise? I couldn't find much help online.

> $ source activate r-peer (r-peer) 
> $ R  
>library(peer) 
> expr <- read.table("~/raw_counts.tsv", header=TRUE)
> dim(expr) 
>[1] 14211   538      # 14211 row-genes, 538 col-samples 
> expr <- t(expr)    # transpose 
> model = PEER()
> PEER_setPhenoMean(model,as.matrix(expr))
> [1] NA                      # in the tutorial, it says this should be "NULL" 
> PEER_setNk(model,10)  # so PEER identifies 10 hidden confounders 
>PEER_getNk(model) [1] 10
> PEER_update(model)
>         iteration 0/1000
>         iteration 1/1000 
>Converged (bound) after 1 iterations NULL     ###### why isn't it iterating?
> 
> 
> sessionInfo() R version 3.4.1 (2017-06-30) Platform:
> x86_64-pc-linux-gnu (64-bit) Running under: Ubuntu 18.04.1 LTS
> 
> Matrix products: default BLAS:
> /home/rodrigo/miniconda2/envs/r-peer/lib/R/lib/libRblas.so LAPACK:
> /home/rodrigo/miniconda2/envs/r-peer/lib/R/lib/libRlapack.so
> 
> locale:  [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C          
> LC_TIME=C.UTF-8  [4] LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8   
> LC_MESSAGES=C.UTF-8  [7] LC_PAPER=C.UTF-8       LC_NAME=C             
> LC_ADDRESS=C [10] LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8
> LC_IDENTIFICATION=C
> 
> attached base packages: [1] stats     graphics  grDevices utils    
> datasets  methods   base
> 
> other attached packages: [1] peer_1.0
> 
> loaded via a namespace (and not attached): [1] compiler_3.4.1

Hi,

I have been going crazy about about this as well. The same problem appeared in the covariates matrix in my case. Basically, if your gene expression dataset is composed of the integers only, then there is a chance that it is read as class integer however the script expects you to input double (Yes, they do not mention this anywhere at all, neither do they provide proper error messages). apply(mat, 2, as.double) resolved the issue with my covariate data.

Best, Farid Ahadli

EDIT: I seriously suspect that this package is written initially for python &| (and or) as a command-line tool. Then their PI asked them to make it available on R as well. In short, very useful but abomination of a package (I do not mean to be disrespectful at all)