_Dockstore_ and _Bioconductor_ for AnVIL
In AnVIL: Bioconductor on the AnVIL compute environment

knitr::opts_chunk$set(
    eval = AnVIL::gcloud_exists(), collapse = TRUE, cache = TRUE
)

Introduction: Basic concepts of Dockstore and Bioconductor

Dockstore is the "VM/Docker sharing infrastructure and management component" of the Global Alliance for Genomics and Health (GA4GH). Dockstore.org implements the infrastructure by defining APIs for coupling Docker images with formalized workflow specifications. The application of this concept to the PanCancer Analysis of Whole Genomes (PCAWG) is described in a 2017 paper by O'Connor and colleagues.

Bioconductor is a software ecosystem based in the R language for the analysis and comprehension of genome-scale experiments. [An overview][] was published in 2015.

Bioconductor implemented a "common workflow" concept a number of years ago. (The term "common" is used on the Bioconductor landing page, to indicate that many bioinformaticians would be expected to engage with tasks reviewed in these workflows. The "common workflow" phrase is not intended to indicate a relationship to CWL, the "common workflow language" standard.)

In Bioconductor parlance, a "common workflow" is an R package accompanied by a markdown or Sweave vignette that narrates and executes the workflow tasks using R. Any sufficiently endowed deployment of R on a sufficient computing platform will run the workflow tasks to completion; this assertion is checked on a regular basis through Bioconductor's continuous integration system. When the assertion fails, the workflow maintainer is alerted and corrections are made.

Advantages to marrying the Bioconductor workflow concept with Dockstore VM/Docker/workflow infrastructure include

Reduction in user burden of configuring and maintaining the execution platform
Utilization of parameterized formal workflow specification in CWL, WDL, or Nextflow
General advantages to users of fostering participation in GA4GH best practices related to reproducibility and transparency

Open questions concern the balance between specification of workflow steps in R and in the formal workflow language. Bioconductor workflows can be written to take advantage of R's capabilities to drive computations on potentially heterogeneous clusters with programmable fault tolerance and job control. The particular advantages of CWL/WDL/Nextflow and other aspects of the Dockstore ecosystem need to be experienced, measured, and documented to help developers establish the appropriate balance between programming R and programming an exogenous workflow environment.

Working with the Dockstore API in Bioconductor

The AnVIL package handles basic aspects of authentication and API element cataloguing for the AnVIL project.

library(AnVIL)
library(dplyr)

Create an object 'dockstore' representing the service and to be used to process API requests.

dockstore <- Dockstore()

Groups of API components are obtained via tags().

knitr::kable(tags(dockstore) %>% count(tag))

We're interested in the 'users' component. Higher level methods will be introduced to help here, but for now we stick to base R methods.

tags(dockstore, "users") %>% print(n = Inf)

We can use the following to determine our user identifier.

myuid <- dockstore$getUser() %>%
    as.list() %>%
    pull("id")

Appendix

Acknowledgments {.unnumbered}

Research reported in this software package was supported by the US National Human Genomics Research Institute of the National Institutes of Health under award number U24HG010263. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.