Skip to content

Auto-apptainer

Auto-apptainer is a help tool that can be used to add software to a SD Desktop virtual machine. It utilizes a library of Apptainer-based software containers, pre-loaded by CSC. You can suggest a container to be added to the library.

Auto-apptainer is not available by default, but you must install it first using SD Software installer

After that you can launch the tool with command:

auto-apptainer

By default, this lists all the Apptainer packages available for auto-apptainer. From the list you can select the package to be installed.

Alternatively you can add a filtering criteria to the command. In this case only those containers whose name or one of commands that the container provides, matches the search criteria, are listed.

For example command:

auto-apptainer bam

asks the user to choose from two containers: bamtools and bedtools. Bedtools is included as it provides commands bamToBed and bamToFasta that match the search criteria in this case.

Once a container is selected, one or more Apptainer wrapper-based commands are created to directory: /shared-directory/sd-tools/bin. These commands can mostly be used like natively installed commands. For example after installing the Bamtools container, Bamtools can be started with command:

bamtools

Below you can find a list of software that can be installed with auto-apptainer. The list can be outdated. You can get an up-to-date list by running command auto-apptainer in SD Desktop.

  • allelecount-4.2.1
  • bamtools_2.5.2--hd03093a_1
  • bcftools_1.20
  • bedtools_2.31.0
  • cnvkit-0.9.12
  • deepvariant_1.5.0
  • EIG_8.0.0
  • hisat2_2.2.1--h87f3376_5
  • htseq-2.0.3
  • image-to-text
  • kneaddata-0.12.0
  • kneaddata-0.12.1
  • mpv-0.34.1
  • p7zio
  • pdf2searchablepdf
  • picard
  • python3-for-medimaging
  • samtools_1.17-2023-06
  • seqtk_v1.3-1-deb_cv1
  • stringtie_2.2.0--ha025227_1
  • taguette.info
  • trimgalore-0.6.9
  • vcftools_v0.1.16-1
  • vsearch_2.23.0
  • weeder_2.0--h9f5acd7_7
Do you find this page useful?
Thank you for your feedback!
Thank you! Your feedback is appreciated.

Please let us know how we could improve this page by contacting us or by creating an issue in GitHub. You can also edit this page yourself and contribute your improvements by creating a pull request.