- Use mosdepth for callability calculations, replacing goleft depth. Centralize
coverage and QC depth calculations around single mosdepth runs.
- Improve representation of germline and somatic calls in MultiQC report and
output directory, avoiding confusing "-germline" extension. Thanks
to Vlad Saveliev.
- Structural variants: return combined tumor/normal calls instead of single
sample tumor for somatic calls in delly, lumpy, manta, and WHAM.
- VarDict: remove `-v 50` as required option for deep targeted panels (>5000x
average coverage). Recommend adding if needed by a `var2vcf` resource options.
- Templating: avoid automatically setting flowcell date to maintain consistency
between runs.
- Add `fusion_caller` as an optional algorithm field to turn on/off fusion
callers. Currently supports oncofuse and pizzly.
- RNA-seq: better appropriate kmer size estimation for reads < 60 bp for
Salmon/Rapmap/Sailfish index creation.
- RNA-seq variant calling: require gatk-haplotype instead of gatk as the caller.
- RNA-seq variant calling: support GATK4
- UMIs: move fgbio consensus calling to use filtering, adds `--max-reads` for
high depth regions and swaps `--min-consensus-base-quality` for `--min-base-quality`
- Correctly re-bgzip fastq inputs even if not using `align_split_size`.
- Fix bug when running with `lumpy_usecnv` that resulted in skipping CNVkit.
- GATK gVCF joint calling: avoid running through bcftools for header fixes,
using Picard instead. Avoids integer/double conversion incompatibilities.
- CWL: run variantcalling with multiple cores, reducing total jobs and enabling
mulicore supporting callers.
- CWL: support structural variant calling as part of variant pipelines.
- Add pizzly (http://www.biorxiv.org/content/early/2017/07/20/166322)
as a fusion caller when fusion mode is enabled.
- VEP: output an effect call per allele for multiallelic positions.
- Define separators for paired fastq files during bcbio_prepare_samples.py
- RNA-seq single-cell/DGE: add `transcriptome_gtf` as an option which will
collapse single-cell/DGE counts down to the gene level. This is recommended
for single-cell and DGE experiments.
- ChIP-seq: preliminary support for bwa for ChIP-seq alignment. Compared to bowtie2
on a test dataset this results in a superset of the bowtie2 peaks, with 95% of the
common peaks within 50 bases of each other. It calls about 50% more peaks
though using the bwa alignments, use with care.