The archs4 package provides utility functions to query and explore the expression profiling data made available through the ARCHS4 project, which is described in the following publication:

Massive mining of publicly available RNA-seq data from human and mouse.

Because this package requires the user to download a number of data files that are external to the package, the installation instructions are a bit more involved than other R packages, and we leave them for the end of this document.

After successful installation of this package, you can query the series and samples included in the ARCHS4 repository, as well as materialize the expresion data into well-known bioconductor assay containers for downstream analysis.

To query GEO series and samples, you can use the sample_info function:

library(archs4)

a4 <- Archs4Repository()
ids <- c('GSE89189', 'GSE29943', "GSM1095128", "GSM1095129", "GSM1095130")
sample.info <- sample_info(a4, ids)
head(sample.info)
#> # A tibble: 6 x 8
#>   series_id sample_id  Sample_title Sample_source_name_ch1 query_type
#>   <chr>     <chr>      <chr>        <chr>                  <chr>     
#> 1 GSE89189  GSM2360252 10318X2      iPS microglia          series    
#> 2 GSE89189  GSM2360253 7028X2       iPS microglia          series    
#> 3 GSE89189  GSM2360254 x2-1         iPS microglia          series    
#> 4 GSE89189  GSM2360255 x2-2         iPS microglia          series    
#> 5 GSE89189  GSM2360256 x2-3         iPS microglia          series    
#> 6 GSE89189  GSM2360257 x2-4         iPS microglia          series    
#> # ... with 3 more variables: sample_h5idx_gene <int>,
#> #   sample_h5idx_transcript <int>, organism <chr>

You can use the as.DGEList function to materialize an edgeR::DGEList from a an arbitrary number of GEO sample and series identifiers. The only restriction is that the data from the series/samples must all be from the same species.

The most often use-case will likely be to create a DGEList for a given study. For instance, the GEO series identifier "GSE89189" refers to the expression data generated to support the Abud et al. iPSC-Derived Human Microglia-like Cells ... paper.

Creating a DGEList from this study will create an object with 27,024 genes across 37 samples in about 1.5 seconds:

yg <- as.DGEList(a4, "GSE89189", feature_type = "gene")

The following command retrieves the 178,135 transcript level counts for this experiment in about 1.5 seconds, as well:

yt <- as.DGEList(a4, "GSE89189", feature_type = "transcript")

The installation of the archs4 package is a bit more involved than a standard package installation and can be roughly broken down into three steps.

1. Install the R package along with its dependencies.
2. Download a number of (large) data files into a specific folder.
3. Generate metadata from the files downloaded in (2) for downstream use.

We will walk you through each step in this section.

The arcsh4 package depends on other packages that are available through both CRAN and Bioconductor. For that reason, we will use the BiocInstaller::biocLite() function to install this package, which can seamlessly install packages from github, CRAN, and Bioconductor.

source("https://bioconductor.org/biocLite.R")
biocLite("denalitherapeutics/archs4", build_vignettes=TRUE)
library("archs4")

When you first load the archs4 library, you will notice a startup message telling you that something isn't quite right with your archs4 installation. The message will look something like this:

Note that your default archs4 data directory is NOT setup correctly

  * Run `archs4_local_data_dir_validate()` to diagnose
  * Refer to the ARCHS4 Data Download section of the archs4 vignette for more information

Your default archs4 data directory (`getOption("archs4.datadir")`) is:

  ~/.archs4data

In order for the package to work correctly, you must download a number of files which are enumerated in the Data File Download section below into a single directory. You will then instruct the archs4 package the path to the directory that holds all of these files by setting the value of R's global "archs4.datadir" option to be the path to that directory.

You will have to create a directory on your filesystem which will hold a number of data files that the archs4 package depends on. Let's call this directory $ARCHS4DIR, which we will define here to be ~/archs4v2data.

The archs4 package provides the archs4_local_data_dir_create() convenience function which creates this directory and copies over a meta.yaml file into that directory. The purpose of this file is to specify the names of the downloaded files that correspond to the human and mouse-level gene and transcript-level data.

library(archs4)
archs4dir <- "~/archs4v2data"
archs4_local_data_dir_create(archs4dir)

Once this directory is created successfully, you will then have to download the following files into it:

• archs4
  • human_matrix.h5: human gene-level counts
  • human_hiseq_transcript_v2.h5: human transcript-level counts
  • mouse_matrix.h5: mouse gene-level counts
  • mouse_hiseq_transcript_v2.h5: mouse transcript-level counts
• ensembl
  • Homo_sapiens.GRCh38.90.gtf.gz: gtf used for human transcript annotations ftp://ftp.ensembl.org/pub/release-90/gtf/homo_sapiens/Homo_sapiens.GRCh38.90.gtf.gz
  • Mus_musculus.GRCm38.90.gtf.gz: gtf used for mouse transcript annotations ftp://ftp.ensembl.org/pub/release-90/gtf/mus_musculus/Mus_musculus.GRCm38.90.gtf.gz

The enumerated items above contain links to the files that need to be downloaded. You can right-click on them and select Save As ... and instruct your web-browser to save them to your local $ARCHS4DIR.

NOTE: Most all of the archs4 functions accept a datadir parameter, which should be the path to $ARCHS4DIR. For convenience, the default value of this parameter is always set to getOption("archs4.datadir"). This means that you can modify your ~/.Rprofile file to set the value of this option to "~/archs4v2data" (for instance), so that the package will always look there by default. If this option is not set in your ~/.Rprofile, the default value for this option is "~/.archs4data".

The datasets currently made available by the ARCHS4 Project only provide minimal feature-level metadata:

• the features in the gene-level datasets are identified only by their symbol; and
• only the ensembl transcript id's are provided for the features in the transcript-level datasets

We want to augment these features with richer annotations, such as the ensembl gene identifiers or gene biotypes, for instance.

To make such data generation automatic and easy for the user, once you have downloaded the Ensembl GTF files listed above into the $ARCHS4DIR, you can run the create_augmented_feature_info() to extract these extra feature-level metadata from the GTF files and store them as tables inside $ARCHS4DIR for later use.

create_augmented_feature_info(archs4dir)

This function will load and parse the GTF files from human and mouse, and create gene- and transcript-level *.csv.gz files in the $ARCHS4DIR which the archs4 package will then later use downstream.

Once your $ARCHS4DIR is setup, you may find it convenient to set the default value for R's global "archs4.datadir" option to the $ARCHS4DIR directory you just setup. To do so, you can put the following line in your ~/.Rprofile file:

options(archs4.datadir = "~/archs4v2data")

Because the installation of this package is a bit more involved than most, we have also provided an archs4_local_data_dir_validate() function, which you can run over your $ARCHS4DIR in order to check on "the health" of your install.

This function will simply look at your $ARCHS4DIR to ensure that the required files are there, and tries to give you helpful error messages if not.

For instance, if the first two files enumerated in the Data File Download section were missing from your $ARCHS4DIR (ie. human_matrix.h5 and human_hiseq_transcript_v2.h5), you would be warned that "something isn't right" when you first load the archs4 package. You could then run the archs4_local_data_dir_validate() to see what is wrong:

archs4_local_data_dir_validate(archs4dir)
#> The following ARCHS4 files are missing, please download them:
#>   * human_matrix.h5: https://s3.amazonaws.com/mssm-seq-matrix/human_matrix.h5
#>   * human_hiseq_transcript_v2.h5: #> https://s3.amazonaws.com/mssm-seq-matrix/human_hiseq_transcript_v2.h5

NOTE: If all installation and data download/processing steps have been completed successfully, a call to archs4_local_data_dir_validate() will simply return TRUE.

If you are developing this package, you will find that it will be convenient to symlink the package's default archs4.datadir path (~/.arcsh4data) to the $ARCHS4DIR you just setup. This is because often times things like roxygen2 document compilation, unit testing, etc. happen in a vanilla R workspace, which won't run the configuration that is prescribed in your ~/.Rprofile file.