broadinstitute/viral-ngs: v1.19.2

New:

• illumina.py illumina_demux now supports simplex runs: non-multiplexed flowcells that do not have a SampleSheet and have no B entries in their read_structure [#776, #759]. Support has also been added for iSeq/FireFly-style flowcell IDs. Simplex-basecalling functionality is not yet tested at the pipeline level (Snakemake or WDL).

Changed:

• Filename cleanups: more assembly WDL tasks now stripping off suffixes left by previous steps (e.g. "taxfilt" or "assembly1-trinity") before appending their new suffixes, resulting in cleaner filenames throughout the assembly workflow. The final refine_2x_and_plot step now creates its final output filename based on the input reads bam instead of the input fasta (as in the pre-WDL dx workflows) and calls the final assembly samplename.fasta (instead of samplename.taxfilt.assembly1-trinity.scaffold.refine2.fasta). [#783]
• Final assembly output file variable is now final_assembly_fasta instead of refine2_assembly_fasta. [#783]

Fixed:

• The WDL refine_2x_and_plot combined task (introduced in the last month as an optimization to reduce instance spin up and staging time) was computing the assembly_length and assembly_length_unambiguous numbers incorrectly: it was computing on the input fasta (from scaffolding), not the output / final assembly. It was also accidentally invoking the plot_coverage python command twice. [#783]
• bmtagger in deplete_human was randomly memory-starved and would fail with "sacrifice child" errors. We have increased bmtagger (srprism)'s RAM limit in the WDL invocation to 90% RAM, up from 50% (since it never runs simultaneously with Java or anything else). [#781, #780]
• Allow Picard SamToFastq more RAM during rmdup_mvicuna_bam [#771]
• bugfix in taxon_filter.py deplete_human backwards compatibility wrapper around deplete [#775]
• bugfix argparse setup for standalone command taxon_filter.py deplete_bwa_bam (was referring to non-existent argument --JVMmemory) [#778]
• No longer silently swallow runtime errors during scaffolding due to a bug in how we parallelized it. [#772]

Added/upgraded:

• upgrade from mummer3 to mummer4, which fixes a few odd bugs, speeds up scaffolding, increases the genome size we can scaffold to, while producing otherwise identical output. [#772, #677]
• bump Picard from 2.13 to 2.17.5 [#774]
• bump Cromwell from v29 to v30.2 [#773]
• bump dxWDL from 0.58.1 to 0.59 [#782]
• bump wdltool from 0.14 to cromwell/womtool 30.2 [#782]
• WDL pipeline: uses GNU Parallel to parallelize up to nproc invocations of single-threaded Krona in the kraken WDL task. Addresses the observation that the Krona for loop was sometimes taking several times longer than Kraken itself. Also parallelizes the single-threaded tar/gzip calls after Krona. [#781]