AnVIL Portal
  • Guides and Tutorials
  • Bioconductor / RStudio
  • Starting RStudio
  • Using R / Bioconductor in AnVIL
  • The R / Bioconductor AnVIL Package
  • Running a Workflow
  • Single-cell RNASeq with 'Orchestrating Single Cell Analysis' in R / Bioconductor
  • Using AnVIL for teaching R / Bioconductor
  • Reproducible Research with AnVILPublish
  • Participant Stories
  • Galaxy
  • Starting Galaxy
  • Jupyter
  • Starting Jupyter

Reproducible Research with AnVILPublish

An exploration of elements of reproducible research with the AnVILPublish package. We will illustrate how to make a docker container tailored for publishing AnVIL packages and then emphasize the merits of an R package structure for organizing research activities in a manner that emphasizes provenance and reproducibility.

Learning Objectives

This week we'll explore elements of reproducible research with the AnVILPublish package. We will illustrate how to make a docker container tailored to a particular purpose (in this case, publishing AnVIL packages!). We'll then emphasize the merits of an R package structure for organizing research activities in a manner that emphasizes provenance and reproducibility. The R package structure coupled with git will form the basis of AnVIL workspace creation, allowing us to maintain a single, version-controlled source for Jupyter notebook or RStudio-based AnVIL workspaces.

Key Resources

Review

Previously...

Essential Steps

  • Login
  • Workspaces
  • Billing accounts
  • (R-based) Jupyter notebooks or RStudio for interactive analysis

Cloud Computing Environment

  • Runtime and persistent disk
  • Workspace DATA and buckets
  • AnVIL package for interaction with workspace components

Workshop Activities

What's the purpose?

  • RStudio provides a rich environment for working in R, but Jupyter notebooks are also relevant, e.g., providing a focused analysis for less-experienced collaborators to walk through.
  • We'd like to be able to provide users with documentation that is accessible in either environment.
  • The documentation should be consistent across environments.

Provenance is important

  • Who wrote or contributed to the software?
  • What does the software do?
  • What license is it available under?
  • What version of the software is currently in use?

Setup

  • Log in to AnVIL using the email address you used to register for the course and navigate (via the HAMBURGER) to Workspaces.
  • Clone the Bioconductor-Workshop-AnVILPublish workspace
    • Unique workspace name
    • Billing project: deeppilots-bioconductor-jun7

Start a CUSTOM CLOUD ENVIRONMENT

  • 'Cloud Environment' in the top right of the workspace, choose 'Customize' Customize Cloud Environment
  • From the 'Application Configuration' dropdown, choose 'Custom Environment' Select Custom Environment
  • For 'Container Image' enter gcr.io/bioconductor-anvil/anvil-rstudio-bioconductor-anvilpublish:3.12-0.0.2 Enter Container Image

R Packages

Create local git clones of the source code of two packages

system2("git", c("clone", "https://github.com/Bioconductor/AnVILPublish"))
system2("git", c("clone", "https://github.com/mtmorgan/AnVILPublishDemo"))

Simple text-based files organize R code, help pages, vignettes, and metadata.

AnVILPublishDemo$ tree
.
├── DESCRIPTION
├── NAMESPACE
├── R
│   └── utilities.R
├── README.md
├── man
│   └── utilities.Rd
└── vignettes
└── A_Introduction.Rmd

Packages are extensible, e.g., all files under an 'inst/' directory are installed with the package

├── inst
│   ├── docker
│   │   ├── anvil-rstudio-bioconductor-anvilpublish
│   │   │   └── Dockerfile
│   │   └── terra-juptyer-bioconductor-anvilpublish
│   │       └── Dockerfile
│   ├── tables
│   │   └── participants.csv
│   └── workflows
│       └── coming_soon.wdl

The DESCRIPTION file provides provenance, including title, version, description, author(s) & their contributions, licensing, as well as system dependencies.

Package: AnVILPublishDemo
Title: Simple Demonstration of AnVILPublish Functionality
Version: 0.0.1
Authors@R:
c(person(
given = "Martin",
family = "Morgan",
role = c("aut", "cre"),
email = "mtmorgan.bioc@gmail.com",
comment = c(ORCID = "0000-0002-5874-8148")
))
Description: AnVILPublish is a way to transform R / Bioconductor
packages, especially vignettes, in Jupyter notebooks for use in
the AnVIL computational environment. The AnVILPublishDemo package
illustrates some of this functionality.
License: Artistic-2.0
Encoding: UTF-8
LazyData: true
Roxygen: list(markdown = TRUE)
RoxygenNote: 7.1.1
Suggests:
knitr,
rmarkdown
VignetteBuilder: knitr

Vignettes

  • A natural place to document what the package does in a narrative 'literate programming' manner. If the code in the vignette does not work, then the package does not build and check successfully.
  • Vignettes may also contain metadata, e.g., the author and date last revised.

From R Package to AnVIL Workspace

Easy!

AnVILPublish::as_workspace(
"~/AnVILPublishDemo",
"deeppilots-bioconductor-jun7",
"AnVILPublishDemo-YOUR_NAME_HERE",
create = TRUE
)

What do we get?

  • DASHBOARD: Provenance -- title, authors, description, version, license
  • NOTEBOOKS: ready to evaluate under an R kernel
  • DATA: tables from packages added. Interpolation of google bucket possible
  • All described in the AnVILPublish vignette

Maybe a little surprising...

  • The package can be developed on your own computer (for instance), and published from there, provided gcloud software is installed.

A Little Under the Hood: Custom Docker Files

Summary

What You've Accomplished

Appreciated R package structure

  • Organizing components of analysis
  • Literate programming
  • Provenance

Transforming packages to notebooks and workspaces

  • Metadata as DASHBOARD entries, including provenance
  • Vignettes as Jupyter notebooks
  • DATA tables populated from the package
  • Coming soon: addition of workflows; creating workspaces for RStudio

Next Steps

Frequently Asked Question

  • What docker images can be used as base images for customization? The main images derive from terra-jupyter-bioconductor (for Jupyter-based images) and anvil-rstudio-bioconductor (for RStudio-based images). Any container can be used in a workflow.
  • Can you specify the runtime environment as part of the workspace? This does not seem to be possible at the moment. One could include a notebook or other code that checked the runtime to see that it meets particular conditions, but this would rely on the user running the code.
  • Enhance reproducibility by 'fixing' package versions, e.g., using packr? Instead, specify precise package versions in a customized Dockerfile.

Help us make these docs great!
All AnVIL docs are open source. See something that’s wrong or unclear? Submit a pull request.
Make a contribution