Skip to content

SYLVER - finding the packages

SYLVER is a publication from the ICR with a coherent description of the code in the Code availability section.

The tutorial takes you through finding where those packages are and building the recipe (the goal is to build an enviroment as specified rather than run all the given code).

Code availablility

As given in the publication code section: 

SYLVER and analysis code was implemented with basic R (v.3.6.0) functionality 
using R packages affy (v.1.64.0), hsentrezgcdf (v.18), 
cdsrmodels (v.0.1.0; R v.4.1.0), limma (v.3.42.2), effsize (v.0.8.1), 
magrittr (v.2.0.1), tidyverse (v.1.3.1), fgsea (v.1.12.0), 
ggplots (v.2_3.3.5), ggrepel (v.0.9.1), org.Hs.eg.db (v.3.10.0), 
VennDiagram (v.1.6.20), survival (v.3.2-11) and GSVA (v.1.34.0). 
SYLVER’s customized code is freely available via Zenodo at 
https://doi.org/10.5281/zenodo.14685952 (ref. 46). 
For details, see relevant sections in Methods.

Methods info

Further info in methods:

DNA methylation data (level 3 β values)—were downloaded from 
http://gdac.broadinstitute.org (release 28 January 2016). 

Statistics and reproducibility
All analyses were performed in R statistical programming environment 
(v.3.6.0, except where stated as v.4.1.0). No data were excluded 
from the analyses unless stated otherwise.

For each TSG, two-class comparison of the genome-wide CRISPR–Cas9 
screen data was performed between these two groups 
(defective and proficient) using linear regression 
(function: cdsrmodels::run_lm_stats_limma) as implemented by the 
Broad Institute’s Cancer Data Science team for these datasets 
(R package: cdsrmodels, v.0.1.0; R v.4.1.0)

Replication steps

1. Generate the initial recipe

Assume you have will use e.g. tutorials/sylver and want to create an inital recipe at sylver.cbl. The recipe template uses the find: directive as we do not kow from the description where the package managers are. 

coble template --recipe tutorials/sylver/sylver.cbl --flavour find
This has creates the cbl file at sylver.cbl with the usual beginning of channels and flags along with a large find section transcribed from the publication: Note that defaults is the higherst priority in the strict channel order due to decommissioned versions that are required by R-3.6.0.

sylver.cbl 
#######################################
# COBLE:Reproducible environment yaml, (c) ICR 2026
#######################################
coble:
  - environment: coble-env
channels:
# note the reverse order of priority  
  - r
  - bioconda
  - conda-forge
  - defaults
languages:
  - r-base=3.6.0@r
flags:
  - dependencies: NA
  - compile-tools: 13.1
  - build-tools: false
  - priority: strict
  - channel: bioconda
  - channel: conda-forge  
find:
  - r-base=3.6.0
  - r-base=4.1.0
  - BiocManager
  - tidyverse=1.3.1
  - effsize=0.8.1
  - magrittr=2.0.1
  - tidyverse=1.3.1
  - ggplot2
  - ggrepel=0.9.1
  - VennDiagram=1.6.20
  - affy=1.64.0
  - fgsea=1.12.0
  - GSVA=1.34.0
  - org.Hs.eg.db=3.10.0
  - survival=3.2-11
  - limma=3.42.2
  - cdsr_models

Note I have already solved some aspects of this environment as the intention is to show the find: directive. Specifically I have fought with channels and channel order.

To resolve these finds simply start to try to build the environment (choose an environment as a name or path, if a path prefix will be used automatically): 

coble build --recipe sylver.cbl --env my-env
This may take some time as the find directove looks for all possible places where the libraries may reside. It will give all the options, so we will need to decide where to take them from.

The coble utility will exit with the message: 

[coble-resolve] Finds were resolved, please check the yml output: tutorials/sylver/sylver.cbl
[*coble] Finds were resolved, please check the yml input before resuming.

It will return IN PLACE an updated cbl file with the best efforts it could make at finding the libraries. NOTE the importance of this as sometimes packages are archived and moved so what is correct at the time of publication may no longer hold at the time of replcaition. Specifically for example in this case R 3.6.0 was current at the time but now no longer on conda-forge.

2. Perfecting the recipe input

This is what we now have:

sylver.cbl 
#######################################
# COBLE:Reproducible environment yaml, (c) ICR 2026
#######################################
coble:
  - environment: coble-env
channels:
# note the reverse order of priority  
  - r
  - bioconda
  - conda-forge
  - defaults

languages:

flags:
  - dependencies: NA
  - compile-tools: 13.1
  - build-tools: false
  - priority: strict
  - channel: bioconda
  - channel: conda-forge  

find:
#   - r-base=3.6.0
found|languages:
  - r-base=3.6.0@r
#   - r-base=4.1.0
found|languages:
  - r-base=4.1.0@conda-forge
#   - BiocManager
found|conda:
  - r-BiocManager@conda-forge
found|conda:
  - r-BiocManager@r
found|r-package:
  - BiocManager@CRAN
found|r-package:
  - BiocManager@CRAN
found|r-package:
  - BiocManager@CRAN

#   - tidyverse=1.3.1
found|conda:
  - r-tidyverse=1.3.1@conda-forge
found|conda:
  - r-tidyverse=1.3.1@r
found|r-package:
  - tidyverse=1.3.1@CRAN
found|pip:
  - tidyverse

#   - effsize=0.8.1
found|conda:
  - r-effsize=0.8.1@conda-forge
found|conda:
  - r-effsize=0.8.1@r
found|r-package:
  - effsize=0.8.1@CRAN

#   - magrittr=2.0.1
found|conda:
  - r-magrittr=2.0.1@conda-forge
found|r-package:
  - magrittr=2.0.1@CRAN

#   - tidyverse=1.3.1
found|conda:
  - r-tidyverse=1.3.1@conda-forge
found|conda:
  - r-tidyverse=1.3.1@r
found|r-package:
  - tidyverse=1.3.1@CRAN
found|pip:
  - tidyverse

#   - ggplot2
found|conda:
  - r-ggplot2@bioconda
found|conda:
  - r-ggplot2@conda-forge
found|conda:
  - r-ggplot2@r
found|r-package:
  - ggplot2@CRAN
found|r-package:
  - ggplot2@CRAN
found|pip:
  - ggplot2@https://github.com/Alicimo/ggplot2,https://github.com/KayRoctaidabee/ggplot2/archive/refs/heads/master.zip

#   - ggrepel=0.9.1
found|conda:
  - r-ggrepel=0.9.1@conda-forge
found|conda:
  - r-ggrepel=0.9.1@r
found|r-package:
  - ggrepel=0.9.1@CRAN

#   - VennDiagram=1.6.20
found|conda:
  - r-VennDiagram=1.6.20@conda-forge
found|r-package:
  - VennDiagram=1.6.20@CRAN
found|r-package:
  - VennDiagram=1.6.20@CRAN
found|pip:
  - VennDiagram@https://github.com/leviperes/VennDiagram,https://github.com/mrForce/VennDiagram,https://github.com/Wayne-Ayers-Creech/VennDiagram/archive/refs/heads/master.zip

#   - affy=1.64.0
found|conda:
  - bioconductor-affy=1.64.0@bioconda
found|bioconductor:
  - affy=1.64.0@r

#   - fgsea=1.12.0
found|conda:
  - bioconductor-fgsea=1.12.0@bioconda
found|bioconductor:
  - fgsea=1.12.0@r

#   - GSVA=1.34.0
found|conda:
  - bioconductor-GSVA=1.34.0@bioconda
found|bioconductor:
  - GSVA=1.34.0@r
found|pip:
  - GSVA

#   - org.Hs.eg.db=3.10.0
found|conda:
  - bioconductor-org.Hs.eg.db=3.10.0@bioconda

#   - survival=3.2-11
found|r-package:
  - survival=3.2-11@CRAN
found|r-package:
  - survival@r-forge
found|pip:
  - survival
found|pip:
  - survival@https://github.com/ryu577/survival/archive/refs/heads/master.zip

#   - limma=3.42.2
found|bioc-package:
  - limma=3.42.2@r
found|pip:
  - limma
found|pip:
  - limma@https://github.com/firoziya/limma/archive/refs/heads/master.zip

#   - cdsr_models

Each section has been commented, and there is a found|packagemanager we can edit it down to something that seems sensible on first effort. We can remove channels and let conda resolve those, and prioritise conda over package managers if we can. The languages sections starts empty, we will move r up to it. We need to add a change to the channel priority after the old version of R has installed.

tutorials/sylver/sylver.cbl 
coble:
  - environment: coble-env
channels:
# note the reverse order of priority  
  - r
  - bioconda
  - conda-forge
  - defaults
languages:
  - r-base=3.6.0@r
  - r-base=4.1.0@conda-forge
bash:
  - conda config --env --add channels bioconda
  - conda config --env --add channels conda-forge
flags:
  - dependencies: NA
  - compile-tools: 13.1
  - build-tools: false
  - priority: strict
  - channel: bioconda
  - channel: conda-forge   
conda:
  - r-BiocManager  
  - r-tidyverse=1.3.1
  - r-effsize=0.8.1
  - r-magrittr=2.0.1
  - r-tidyverse=1.3.1
  - r-ggplot2
  - r-ggrepel=0.9.1
  - r-VennDiagram=1.6.20
  - bioconductor-affy=1.64.0
  - bioconductor-fgsea=1.12.0
  - bioconductor-GSVA=1.34.0
  - bioconductor-org.Hs.eg.db=3.10.0
r-package:
  - survival=3.2-11
bioc-package:
  - limma=3.42.2

#   - cdsr_models

We can see that cdsr_models was not found, and I can neaten the conda installs to specify biocmanager and R. I will remove cdsr_models for this tutorial (further explanation coming). The automatic prepend of r and bioconductor facilitates moving the packages around for troubleshooting or version change. Because 3.6.0 is quite old we need some flexibility so we move the order around in the flags.

tutorials/sylver/sylver.cbl 
coble:
  - environment: coble-env
channels:
# note the reverse order of priority  
  - r
  - bioconda
  - conda-forge
  - defaults
languages:
  - r-base=3.6.0@r
  - r-base=4.1.0@conda-forge
flags:
  - dependencies: NA
  - compile-tools: 13.1
  - build-tools: false
  - priority: strict
  - channel: bioconda
  - channel: conda-forge     
r-conda:
  - BiocManager  
  - remotes # we need remotes for installs
  - tidyverse=1.3.1
  - effsize=0.8.1
  - magrittr=2.0.1
  - tidyverse=1.3.1
  - ggplot2
  - ggrepel=0.9.1
  - VennDiagram=1.6.20
bioc-conda:
  - affy=1.64.0
  - fgsea=1.12.0
  - GSVA=1.34.0
  - org.Hs.eg.db=3.10.0
r-package:
  - survival=3.2-11
bioc-package:
  - limma=3.42.2

3. Try updated version

Put the above recipe in the sylver.cbl file and then run again with the same command. 

coble build --recipe tutorials/sylver/sylver.cbl --env my-env
It fails :-) with the message 
Error: More than one R version specified in languages section.
[*coble] Error in creating recipe, please correct input before resuming.
This is the defined behaviour as we cannot have 2 different versions of R in the conda environment. Looking back at the publication, 4.1.0 is only used for cdsr_models, which we removed. Now we see it would need a seperate environment, so we will leave this for now and continue with just the 3.6.0 environment - update and re run:

# Optionally pull in the ready-completed template
coble template --recipe tutorials/sylver/sylver.cbl --flavour sylver
# Then build
coble build --recipe tutorials/sylver/sylver.cbl --env my-env
Final version: tutorials/sylver/sylver.cbl 
#######################################
# COBLE:Reproducible environment yaml, (c) ICR 2026
#######################################
coble:
  - environment: coble-env
channels:
# note the reverse order of priority  
  - r
  - bioconda
  - conda-forge
  - defaults
languages:
  - r-base=3.6.0@r
flags:
  - dependencies: NA
  - compile-tools: 13.1
  - build-tools: false
  - priority: strict
  - channel: bioconda
  - channel: conda-forge  
r-conda:
  - BiocManager
  - remotes
  - tidyverse=1.3.1
  - effsize=0.8.1
  - magrittr=2.0.1
  - tidyverse=1.3.1
  - ggplot2
  - ggrepel=0.9.1
  - VennDiagram=1.6.20
bioc-conda:
  - affy=1.64.0
  - fgsea=1.12.0
  - GSVA=1.34.0
  - org.Hs.eg.db=3.10.0
r-package:
  - survival=3.2-11
bioc-package:
  - limma=3.42.2

4. Tracking the output logs

Now we are running we can track the logs.
- sylver.sh - the cbl transformed into a pure bash script that could be run instead
- sylver.log - each bash install cleans and runs here so you can track the stdout
- sylver.err - each bash line cleans the err file so you can track the current stderr
- sylver_summary.txt - after each install the logs are parsed for important info eg errors or dependencies. This is output along with the timings so you can keep an eye on the entire install here.