Skip to content

0.109.0

Choose a tag to compare

@derijkp derijkp released this 07 May 15:32
· 218 commits to master since this release

Major changes are:

Qol improvements

  • added cg project_make for creating a basic experiment/project directory based on a samplesheet
  • Added option -samplesheet to cg_process_project (uses project_make internally) (issue derijkp/scywalker#3)
  • Improvement help
  • Added cg error_report for easier checking of errors after a distributed run (issue derijkp/scywalker#5 and derijkp/scywalker#4)

Distribution and installation

  • Added cg install for easier install of supporting binaries and refdbs
  • include more dependencies in distro (from testing in minimal docker distro)
    • less, which, bzcat, bzip2, dot
    • extra libs for dirR
    • semi-static compiles of samtools, bgzip, tabix

Updates of the reference databases (hg38)

  • dbsnp 155
  • gnomad 4.1
  • gencode 45, ...

more cram support:

  • cram can be used with ont presets (use -aliformat cram)
    • fix remora: modkit has problems with cram -> temp convert to bam to run
    • scywalker_report_job: support cram
  • cg tsv2sam now supports -outformat cram

falco replaces fastqc
falco is a (mostly) drop-in replacement that is faster, smaller and comes with less dependencies than fastqc

Optimization cg select summaries

  • by default making summaries (-g option) memory is proportional to size of output
    (other queries are independent of size)
  • added option -optim memory
    • minimize memory use when making summaries with option -g
    • Using the option is slower, so do not use when (typically) small summary output is expected
    • used in scywalker for getting aggregate data from sc_gene and sc_isoforms

Optimizations (esp. for starting big runs)

  • file access optimizations (check for compressed only where specified, direct deps check, use cached data)
  • storage optimizations (shadowdirs)
  • jobs optimizations (integrate cleanup, merge jobs, hold_jid in runfile)

Other changes

  • Added extra cluster support (options -submitoptions, -dqueue, help, ...) (issue #1)
  • process_project option -jobsample now allows specification of number of cores to be used by one-job sample analysis
  • cg_cat: added option -fullfilename
  • Changed preset_ont: removed npinv from default svcallers (as it often caused problems with newer data)
  • cg fixtsv will also fix duplicate fields in tsv file

Full Changelog: 0.108.0...0.109.0