Taming Snakemake
Download as PPTX, PDF5 likes4,077 views
This document discusses the advantages of using Make and Snakemake for bioinformatics workflows compared to other options like shell scripts and Perl. It outlines key features of Make like its focus on files, ability to restart safely, and support for parallelization. The document then introduces Snakemake as building upon Make's strengths while addressing weaknesses like lack of control flow and cluster support. It provides examples of using Snakemake with cluster resources and utilities like logs, dry-runs and workflow diagrams. Finally, it discusses best practices like versioning, caching, integration with R, leaving paper trails and ensuring reproducibility.
Taming Snakemake
- 2. Why Make? What are Make's advantages (over Perl and shell scripts)? Make forces you to think about file transformation in terms of inputs and outputs, recipes and rules. In Perl you are forced to think at the level of variables, conditionals, and loops. In Shell you are forced to think like a caveman. Unfortunately, bioinformatics is still largely about files and their suffixes. Make has a very powerful syntax based almost entirely around file suffixes. Make knows what's been made and what hasn't. Make can be interrupted and restarted safely, and without overwriting finished work. Make knows what's changed and what hasn't. If an input is newer than an output, it will attempt to rebuild the output. Make allows you to add new input files without worrying about overwriting old ones. Make is well supported. There are 1333 Make questions on SO alone. When people see a Makefile, they immediately know how to run it. Make does not force you to wrap shell statements in quotes. Make is a DSL. It will attempt to validate your syntax. Make is ancient, ubiquitous, and reliable. Make can parallelize with --jobs. Make recipes encourage reuse. https://share.chop.edu/pages/viewpage.action?pageId=138478819
- 3. Make review http://github.research.chop.edu/BiG/err_chip_seq/blob/master/Makefile
- 6. Why Snakemake? Addresses Makefile weaknesses without throwing out the good stuff Difficult to implement control flow No cluster support Inflexible wildcards Too much reliance on sentinal files No reporting mechanism Johannes Köster
- 8. Utilities Logs - wire them up manually Cluster support pretty decent source /nas/is1/leipzig/martin/variome-env/bin/activate snakemake --directory /nas/is1/leipzig/martin/snake-env --snakefile /nas/is1/leipzig/martin/snake-env/Snakefile -c qsub -j 16 source /mnt/isilon/cbmi/variome/leipzig/martin/respublica-env/bin/activate snakemake --directory /mnt/isilon/cbmi/variome/leipzig/martin/snake-env --snakefile /mnt/isilon/cbmi/variome/leipzig/martin/snake-env/Snakefile -c qsub -j 16 Cores/jobs/resources
- 9. Useful stuff dry-runs keep-going touch version changes workflow diagrams
- 10. Python legal
- 11. Client websites with Jekyll Jekyll is a templating engine for blogs that accepts Markdown Layouts use the Liquid markup http://mitomap.org/martin-rnaseq/
- 12. A workflow that reports itself
- 14. The bad way
- 18. R/Snakemake integration git submodule add git@github.research.chop.edu:BiG/rna-seq-common-functions.git common/rna-seq
- 19. Leave a paper trail
- 20. Reproducible Checklist repository github.research.chop.edu workflow of some kind from beginning to end website at mybic.chop.edu
- 21. Ties that bind