Snakemake with R script, error: snakemake object not found

Question

I'm having issues trying to run a snakemake rule that invokes an R script. The relevant snakemake rule looks like this:

rule summarize_transcripts:
    input:
        lineages = expand("../final/{seq_run}/{exp}_transcript_lineages.csv", exp=EXP, seq_run=SEQ_RUN),
        salmon_dir = expand("../salmon_quant_results/{experiment_id}", experiment_id=EXP),
    output:
        species_counts = expand("../final/{seq_run}/{exp}_sp_counts.csv", exp=EXP, seq_run=SEQ_RUN),
        family_counts = expand("../final/{seq_run}/{exp}_family_counts.csv", exp=EXP, seq_run=SEQ_RUN)        
    shell:
        '''
        Rscript ./deseq2.R
        '''

Here's the top of my R script, which fails at the S4 object call:

library('DESeq2')
library('tximport')

#Read in dataframe that contains the trinity ID's linked to lineage information
df <- read.csv(snakemake@input[["lineages"]], header=TRUE)

The libraries load fine, the error message states:

Error in read.table(file = file, header = header, sep = sep, quote = quote,  : 
  object 'snakemake' not found
Calls: read.csv -> read.table
Execution halted

I'm trying to keep my dependencies organized within a conda .yml file, which looks like this:

name: test_env

channels:
  - conda-forge
  - bioconda
  - r
  - default

dependencies:
  - star =2.7.1a
  - trinity =2.12.0
  - samtools =1.12
  - salmon =1.4.0
  - snakemake
  - pandas
  - r=4.1.0
  - r-essentials
  - bioconductor-deseq2
  - bioconductor-tximport

Why is R not recognizing the snakemake inputs as described in the snakemake docs? I am running snakemake version 6.5.1.

Answer 1

The examples use script: rather than shell: So change to

script:
  "./deseq2.R"

rather than

shell:
    '''
    Rscript ./deseq2.R
    '''

It looks like snakemake needs to inject R code into the script to define the snakemake variable and it cannot do that with arbitrary shell commands. This is only done when the script file ends with ".R"