Repository navigation
Tool Versions
Every tool version is pinned independently with defaults shipped in
workflow/default-config/versions.yaml (STAR 2.7.11b, samtools 1.22.1,
TrimGalore! 0.6.10, FastQC 0.12.1, RSeQC 5.0.4, MultiQC 1.33, TEtranscripts
2.2.4, TElocal 1.1.3, StringTie 2.2.1, DESeq2 1.46.0, pheatmap 1.0.12,
UCSC tools 482). Override any by creating
input/versions.yaml with just the keys you want to change:
versions:
star: "2.7.10b"The tools are installed into the shared conda env
(workflow/environment.yaml) at env creation, with the same pins as
versions.yaml — keep the two files in sync when bumping a version, then
conda env update -f workflow/environment.yaml to apply. The pre-built
environment tarball on GitHub Releases is built from this same file in CI, so
it carries the identical pins. Two deliberate exceptions: samtools is pinned
to 1.22.1 because STAR 2.7.11b links htslib <1.23, incompatible with
samtools 1.23+ in one environment; and TEtranscripts is pip-installed
(TEtranscripts==2.2.4) because the bioconda recipe pins an ancient
bioconductor-deseq (DESeq v1) that can only coexist with R 4.0-era packages,
making it unsolvable alongside a modern DESeq2 (details in
workflow/environment.yaml). As a fallback, --sdm conda is still supported:
Snakemake then generates one small per-tool env per rule from versions.yaml
at parse time instead of using the shared env.
If STAR/samtools/TrimGalore!/FastQC/RSeQC/MultiQC/TEtranscripts are already
on your PATH (e.g. via module load on a cluster), just run snakemake
without the shared env activated — Snakemake runs the plain shell commands.
You then own the version matching; the shared conda env exists to guarantee
it automatically.