Do not pass pandas object to count script as divergent pandas version… (

#12)

* Do not pass pandas object to count script as divergent pandas versions between host process and script might cause problems with deserialization.

* do not check names when reading sample sheet

* pin star version when creating index

* plain python list and strings

* fixes

* remove debug output

.travis.yml

@@ -16,7 +16,7 @@ install:
   - conda config --add channels conda-forge
   - conda config --add channels bioconda
   - conda create -q -n snakemake snakemake>=5.1.2 python=$TRAVIS_PYTHON_VERSION
-  - conda create -q -n star star
+  - conda create -q -n star star=2.5.3a
   - source activate star
   # create STAR index (too big to live in git)
   - mkdir .test/data/ref/index

rules/diffexp.smk

@@ -4,7 +4,7 @@ rule count_matrix:
     output:
         "counts/all.tsv"
     params:
-        units=units
+        samples=units["sample"].tolist()
     conda:
         "../envs/pandas.yaml"
     script:

scripts/count-matrix.py

@@ -3,12 +3,11 @@
 counts = [pd.read_table(f, index_col=0, usecols=[0, 1], header=None, skiprows=4)
           for f in snakemake.input]
 
-for t, (sample, unit) in zip(counts, snakemake.params.units.index):
+for t, sample in zip(counts, snakemake.params.samples):
     t.columns = [sample]
 
 matrix = pd.concat(counts, axis=1)
 matrix.index.name = "gene"
 # collapse technical replicates
 matrix = matrix.groupby(matrix.columns, axis=1).sum()
-print(matrix)
 matrix.to_csv(snakemake.output[0], sep="\t")

scripts/deseq2-init.R

@@ -14,9 +14,8 @@ if (snakemake@threads > 1) {
 
 # colData and countData must have the same sample order, but this is ensured
 # by the way we create the count matrix
-cts <- read.table(snakemake@input[["counts"]], header=TRUE, row.names="gene")
-coldata <- read.table(snakemake@params[["samples"]], header=TRUE, row.names="sample")
-print(cts)
+cts <- read.table(snakemake@input[["counts"]], header=TRUE, row.names="gene", check.names=FALSE)
+coldata <- read.table(snakemake@params[["samples"]], header=TRUE, row.names="sample", check.names=FALSE)
 
 dds <- DESeqDataSetFromMatrix(countData=cts,
                               colData=coldata,