dib-lab / elvers

(formerly eelpond) an automated RNA-Seq workflow system
https://dib-lab.github.io/elvers/
Other
28 stars 3 forks source link

fix deseq2 conda env #140

Closed bluegenes closed 5 years ago

bluegenes commented 5 years ago

Log: (starting after pkg install info)

reading in files with read.delim (install 'readr' package for speed up)
using counts and average transcript lengths from tximport
estimating size factors
using 'avgTxLength' from assays(dds), correcting for library size
estimating dispersions
Error in checkForExperimentalReplicates(object, modelMatrix) :

  The design matrix has the same number of samples and coefficients to fit,
  so estimation of dispersion is not possible. Treating samples
  as replicates was deprecated in v1.20 and no longer supported since v1.22.

Calls: DESeq ... estimateDispersions -> .local -> checkForExperimentalReplicates
Execution halted
bluegenes commented 5 years ago

environment problems solved, but the command elvers examples/yeast.yaml quantify diffexp -n still having some trouble (failing randomly). I think what happens is that snakemake finds the targets for the diffexp workflow first, which causes the "sample" wildcard to be read wrong:

KeyError: 'ERR458500_1'
Wildcards:
sample=ERR458500_1
assembly=yeast

where sample should be sample=ERR458500, with 1 being the unit value.

Error only occurs when either the diffexp target or the deseq2 target is included.

bluegenes commented 5 years ago

Ok, this issue is actually being brought on by combining units at the salmon step (combine all files per sample for salmon mapping). The issue was only showing up with samples with only a single unit value (fake units). I've put a placeholder "ignore_units" in place to ignore "combining" if there's any single-unit samples, but this isn't a long-term fix. I think it'd be better to combine units at the "get_data" stage, and never consider "units" in any downstream steps. I have an old branch to work on this and will revive the effort.

bluegenes commented 5 years ago

..and we're back to this genomeinfoDb pkg install issue (see https://github.com/bioconda/bioconda-recipes/issues/13846). Updating conda to 4.6.12 worked on my ubuntu 18.04 instance, so trying that in the travis yml file.

conda update -n base conda
conda clean --all # to remove older genomeinfodb pkg