Skip to contents

Working with isolates and fetching data from SRA

Source: vignettes/sra_fetching.Rmd
sra_fetching.Rmd

When to use this

Zorn/Bascet can also work with isolates, treating each isolate as a “cell”, giving access to all clustering commands etc. This can be handy if you have a larger number of genomes as you won’t drown you file system with FASTA files which reduces performance.

This workflow is in its infancy and currently works best with raw reads. You can likely download assembled genomes as well but some steps need rehashing to ensure that you don’t run into problems.

Input via SRA

Zorn/Bascet is able to download FASTQ files from SRA, packing the data directly into TIRP files (i.e., as if you had run single-cell and just debarcoded the data). The SRA workflow also downloads metadata per file that you can later use to rename the cells. Initially, cells are named by the run, i.e., SRRxxxx.

Requirements for SRA-tools

Zorn uses the rentrez R package to query NCBI. Bascet uses SRA Toolkit for the actual downloads, so prefetch and fasterq-dump must be available on the machine that runs the download.

Install SRA Toolkit before running BascetDownloadSraRuns() or bascet import-sra. BascetDownloadSraRuns() checks that both executables can be found before submitting the Bascet jobs.

Download and install SRA Toolkit from NCBI: https://github.com/ncbi/sra-tools/wiki/01.-Downloading-SRA-Toolkit. After installation, check that the tools are on PATH:

which prefetch
which fasterq-dump

If the tools are not on PATH, you can pass their full paths:

BascetDownloadSraRuns(
  bascetRoot = "/path/to/bascet_root",
  prefetch = "/path/to/prefetch",
  fasterqDump = "/path/to/fasterq-dump"
)

Downloading SRA runs

The scMetaG study by Zheng et al. 2022 deposted each individual cell as a single file. This makes the files hard to download manually (the SRA run selector struggles to even open if you send all the files to it!). But Zorn/Bascet can take the ID of their deposition, PRJNA803937, and first fetches the list of runs:

#Get a list of SRRs and metadata
tab <- BascetResolveSra(
  bioproject = "PRJNA803937",
  terms = "SAG_"  #only SAGs; adjust for your needs
)

#Write shards of SRRs to be fetched
BascetPrepareSraFetchLists(
  tab,
  bascetRoot = bascetRoot
  #Default is 1000 runs per TIRP
)

This creates “sralist”-files holding a list of SRAs to download into each TIRP. It also makes a runinfo.csv-file, holding metadata per run. The latter you can use to add metadata at a later stage (optional).

Then download all shards into TIRP files:

#Download the SRA runs
BascetDownloadSraRuns(
  bascetRoot = bascetRoot,
  threads = 16
  #runsAhead = 10   #optional to set. currently each thread will download a separate SRA-run, but you can lower this if the runs are large
)

From here, you can proceed as if had just debarcoded reads!