The R / Bioconductor AnVIL Package

An exploration of how workspaces provide a framework for managing data and large-scale analyses using the HCA Optimus Pipeline and 1000G-high-coverage-2019 workspaces and R using the AnVIL package.

Learning Objectives

This week we'll explore how workspaces provide a framework for managing data and large-scale analyses. We use the HCA Optimus Pipeline and 1000G-high-coverage-2019 package.

Key Resources

• Visit https://anvil.terra.bio to use the AnVIL platform.
• We use week-2-demo.R to guide us through this workshop.
• We use the HCA Optimus Pipeline and 1000G-high-coverage-2019 workspaces as examples.
• Review the Introduction to the AnVIL package vignette.

Review

Previously...

• Notes and recorded session: Using R / Bioconductor in AnVIL

Essential Steps

• Login
• Workspaces
• Billing accounts
• Cloud environment -- (R-based) Jupyter notebooks or RStudio

Cloud Computing Environment

• Runtime and persistent disk
• A 'personal' cloud computing environment
• Not shared with others
• Ephemeral

FAQs

• Persistent disk mounted at
  • R / Jupyter: /home/jupyter-user/notebooks
  • RStudio: /home/rstudio
• Startup script or custom docker file for 'sudo'-like access, and for complete reproducibility

Workshop Activities

Setup

• Log in to AnVIL using the email address you used to register for the course and navigate (via the HAMBURGER) to Workspaces.
• If you cloned the Bioconductor-Workshop-Popup workspace last week, delete it now.
• Clone the Bioconductor-Workshop-Popup.
• Start an RStudio cloud environment.
• Launch the cloud environment.
• Copy the week-2-demo.R script into a file on your cloud environment.

Workflows

• In a new browser tab/window, navigate (via the HAMBURGER) to the HCA Optimus Pipeline workspace. This workspace demonstrates how scRNA-seq fastq files can be transformed to a 'count matrix' for interactive analysis.
• Overall orientation: DATA TABLES serve as input to WORKFLOWS (scalable 'big data' computation).
• Workflows transform big data using 'Workflow Description Language' scripts producing outputs (logs, results). For this workflow:
  • Single-cell RNA seq analysis.
  • Inputs are fastq files from individual samples.
  • Scripts perform alignment, UMI processing, creating a 'count' matrix of gene x cell (sample) expression matrices, etc.
  • Primary output of interest is a 'loom' file summarizing the count matrix.
• Workspace bucket / Files store workflow outputs (each workflow run has a unique identifier; logs and results are located under the identifier). Buckets also provide a location for storing and sharing interactive analysis results.

The AnVIL Package

AnVIL Workspaces

hca = "featured-workspaces-hca/HCA_Optimus_Pipeline"
thousand_genomes = "anvil-datastorage/1000G-high-coverage-2019"

library(AnVIL)
avworkspace()    # current workspace
avworkspace(hca) # set to HCA workspace

DATA TABLE Access

avtables()

tbl = avtable("sample")
tbl

tbl %>% count(participant)

## tbl %>% avtable_import()

avworkspace(thousand_genomes)
avtables()
participant = avtable("participant")
participant

participant %>% count(POPULATION, sort = TRUE)
avtable("pedigree") %>%
    count(Population, Sex) %>%
    tidyr::pivot_wider(names_from = "Sex", values_from = "n")

## switch back to this workspace
avworkspace(hca)

Google buckets

## Copy files from google buckets to persistent disk

tbl = avtable("sample_set")
tbl

dir.create("~/loom")
gsutil_cp(tbl$loom_output_file, "~/loom/")  # see also gsutil_rsync()
dir("~/loom")

## Workspace Bucket -- 'backup' or share persistent disk to workspace bucket

avbucket()  # bucket associated with this workspace
gsutil_ls(avbucket())

avfiles_backup("~/scripts", recursive = TRUE) # see also avfiles_restore()
gsutil_ls(avbucket(), recursive = TRUE)

Fast Binary Package Installation

## do NOT update out-of-date packages yet
BiocManager::install("Bioconductor/AnVIL")

## RESTART R
AnVIL::repositories() # binary Bioconductor and CRAN package installation

## install and use LoomExperiment
AnVIL::install("LoomExperiment") # about 40 seconds, rather than 10's of minutes
sce = LoomExperiment::import("~/loom/pbmc_human_v3.loom")

Access AnVIL from Outside AnVIL

• Requires gcloud SDK installed on your computer.
• Use SDK to register your Gmail account and google billing project.

Access the AnVIL 'API'

leo = Leonardo()
leo
leo$listDisks()

terra = Terra()
tags(terra, "Workspaces")
wkspc =
    terra$listWorkspaces() %>%
    flatten() %>%
    select(-starts_with("workspace.attributes"))
wkspc

Summary

What You've Accomplished

Setup

• Clone a workspace, launch an RStudio cloud environment
• Navigate between workspaces

Workflows

• Elements of workflow structure -- DATA TABLE inputs, scripts, File outputs

AnVIL Package

• Selecting workspaces
• Managing DATA TABLEs
• Moving data to and from google buckets
• Fast binary package installation (in the 'devel' version of the package)
• Advanced features, e.g., local use, API access

Next Steps

• Follow instructions at Set up billing with $300 Google credits to explore Terra to enable billing for your own projects.

Frequently Asked Questions

• Uploading workflows -- through GitHub / Dockstore, but also the Broad Methods Repository (YouTube); see also the WDL Puzzles workspace.
• Default name and namespace -- the runtime starts in a particular workspace, and the runtime knows the default namespace and name. So by default, I had

  > avworkspace()
  [1] "deeppilots-bioconductor-may3/Bioconductor-Workshop-PopUp-mtmorgan"
  

• gsutil_cp(): CommandException: Downloading this composite object requires integrity checking with CRC32c, but your crcmod installation isn’t using... This is a bug that should be fixed in the underlying image for the runtime.