HISTORY.md

1.1.1 (in progress)

• tumor-only prioritization: do not apply LowPriority filter by default, instead annotate with external databases. Use tumoronly_germline_filter to re-enable previous behavior.
• UMIs: apply default filtering based on de-duplicated read depth. Uses --min-reads 2 with raw de-duplicated coverage of 800 or more or --min-reads 1 otherwise. Allows error correction with UMIs for higher depth samples.
• gemini: databases no longer created by default. Use tools_on: [gemini] or tools_on: [gemini_orig] to create a database.
• vcfanno: run gemini and somatic annotations by default, producing annotated VCFs with external information.
• alignment preparation: support a list of split files from multiple sequencing lanes, merging into a single fastq
• variant: support octopus variant caller for germline and somatic samples.
• peddy: fix bug where not all files uploaded on first pipeline run
• peddy: skip runs on somatic analyses where we do not get reliable output
• GATK: support ploidy specification during joint calling.
• GATK BQSR: bin qualities into static groups (10, 20, 30) to match GATK4 recommendations. Thanks to Severine Catreux.
• variant calling: support bcftools 1.9 which is more strict about duplicated key names in INFO and FORMAT.
• seq2c: Upload global calls, coverage and read_mapping files to project directory.
• RNA-seq variant calling: Apply annotations after joint calling for GATK to avoid import errors with GenomicsDB. Thanks to Komal Rathi.
• CWL: add --cwl target to bcbio_nextgen.py upgrade to add and maintain bcbio-vm.
• CWL: use standard null instead of string "null" for representing None values.
• CWL: support for heterogeneity and structural variant callers that make use of variant inputs.
• CWL: support ensemble calling for combining multiple variant callers.
• RNA-seq: output a matrix of un-deduped UMI counts when doing single-cell/DGE for quality control purposes. This is called tagcounts-dupes.mtx in the final directory.
• single-cell RNA-seq: allow pre-transformed FASTQ files as input to DGE/single-cell pipeline.
• single-cell RNA-seq: only create one index per specified genome instead of per sample
• fgbio: back compatibility for older quality setting --min-consensus-base-quality
• RNA-seq: fix for fusion_caller getting interpreted as a path, leading to memoization/upload issues.
• RNA-seq: memoize rRNA quality calculations, speeding up reruns.
• RNA-seq: prefix description with an X if it starts with a number, for R compatibility. Thanks to Avinash Reddy and Dan Stetson at AstraZeneca.

1.1.0 (11 July 2018)

• Germline calls: rename outputs to samplename-germline to provide easier to understand outputs in final directory.
• Add bcbioRNASeq object creation and automatic quality report generation with tools_on: [bcbiornaseq]
• CWL: Support germline/somatic calling for tumor samples.
• CNVkit: improve whole genome runs. Better speed in normalize_sv_coverage through parallelization and avoiding logging. Avoid memory errors in segmentation.
• UMI: upload prepared UMI bam file (pre-consensus) to final output directory
• Add support for bbmap as an aligner
• RNA-seq variant calling: parallelize GATK HaplotypeCaller over regions to avoid memory and timeout issues.
• Support joint calling with GATK using pre-prepared gVCF inputs.
• RNA-seq variant calling: allow annotation of output variants with vcfanno
• Support hg38 builds with peddy QC
• QC: support VerifyBamID2 for contamination detection
• CWL: adjust defaults for align_split_size and nomap_split_targets to match different parallelization and overhead for these runs
• CWL: support for Cromwell runner
• custom genomes: Unzip GTF file prior to installation.
• Avoid making variant_regions required during processing (by filling with coverage) to differentiate targeted and non analyses downstream.
• Avoid attempts to download pre-installed S3 genomes, providing better errors with missing genome installs.
• Trimming: add explicit polyg option for removing 3' G stretches in NovaSeq and NextSeq data. Now defaults to no polyG trimming unless turned on.
• Chip-seq: Add RiP calculation for chip-seq data.
• DeepVariant and Strelka2 support for customized targeted/genome calling models per region to handle heterogeneous inputs.
• STAR: enable passing custom options for alignment.
• Add tools_off: [coverage_qc] option to skip calculating coverage stats (samtools-stats and picard).
• Adding BAM file for each sample in small-RNAseq pipeline, samtools and qualimap qc metrics to multiqc report.
• Allow arbitrary genomes for ChIP-seq. Thanks to @evchambers for pointing out the issue.

1.0.9 (10 April 2018)

• Use smoove for lumpy variant calling and genotyping, replacing custom lumpyexpress implementation: validation
• Generalize exclusion of regions during variant calling with new exclude_regions target. Includes previously available LCR and high depth regions, in addition to removal of polyX and alternative contigs.
• Normalize allele frequency calculation and filtering for Strelka2 and MuTect2. Thanks to Vlad Saveliev.
• CNVkit: enable specification of pre-built reference background cnn with background: cnv_reference.
• CNVkit: handle projects with mixed CNVkit and non-CNVkit usage. Thanks to Luca Beltrame.
• Improved Atropos trimming: better use of multicore parallelization in variant and RNA-seq pipelines.
• Add support for polyG and polyX trimming to variant calling for NovaSeq 3' end cleanup and generally avoiding low complexity reads.
• Structural variant: use SURVIVOR for validation comparisons.
• RNA-seq variant calling: use multiple cores for VarDict.
• Support miRge2.0 for alternative small RNA annotation. Users should install the tool manually until compatible with bioconda.
• Add bamCoverage to chip-seq pipeline to calculate bigwig files.
• GATK4: Correctly use GATK4 GatherVcfs when tools_off: [gatk4] specified for variant calling. Thanks to Luca Beltrame.
• variant: Default to mark_duplicates: false if alignment turned off (aligner: false).
• variant: Fix race condition when preparing BED files for coverage and sv_regions. Thanks to Tristan Lubinski.
• Fix noalt_calling to correctly avoid parallelizing on non-standard chromosomes without a variant regions file.
• Fix broken kraken command. Thanks to @choosehappy.

1.0.8 (5 February 2018)

• GATK4 is the new default GATK release used in bcbio when running HaplotypeCaller or Base Quality Score Recalibration. Use tools_off: [gatk4] to use older GATK 3.x versions.
• GATK4: Support 4.0 release with changed command line parameters. Re-enable multicore calling for CWL runs.
• GATK4: remove older GATK3 based gatk-framework in favor of equivalent GATK4 commands.
• install: move to using recent IPython parallel to avoid dependency issues
• install: fix resolution issues due to conda 4.4.x (old ipython-cluster-helper, missing libquadmath.so with numpy due to libgcc update)
• RNA-seq variant calling: improve joint calling and parallelization with move to use GATK4 HaplotypeCaller.
• QC: improve read counting speed by moving to hts-nim-tools, replacing custom samtools view counting
• Add noalt_calling to avoid variant calling on non standard chromosomes. Thanks to Vlad Saveliev and Oliver Hofmann.
• variant alignment: improve core allocations for non-split alignments to avoid memory issues on 4Gb/core machines with whole genome samples.
• Add Total number of reads and adapter found to metrics in small RNA-seq pipeline.
• Add mirtop to the tools used in small RNA-seq pipeline for miRNA annotation.
• delly: Support 0.7.8 release which calls all variant types together.
• gVCF: fix basic filtering for GATK and sentieon when running without joint calling. Thanks to Tom Morris.

1.0.7 (6 January 2018)

• Automatically include bcbio anaconda PATH when running tools. Also allow custom BCBIOPATH specification to help with modules integration. Thanks to Gabriel Berriz.
• vcfanno: only correct VCF headers to use Number=1 when decomposition takes place. Avoids incorrect headers for non-decomposed inputs.
• ensemble: normalize and decompose variants prior to incorporating into ensemble calls, handling MNPs called differently across callers. Thanks to Vlad Saveliev.
• Avoid bgzipping and grabix indexing fastq inputs when not doing alignment splitting to save processing.
• Initial support for minimap2 aligner in variant calling workflows. Still needs validation and benchmarking in comparison to bwa.
• Standardize dbSNP annotation to use vcfanno for all variant callers. Remove GATK custom annotations for non-GATK callers, which are not present in GATK4.
• CNVkit: drop low coverage contaminating regions in tumor calls. Thanks to Eric Talevich.
• Expand remove_extracontigs for bam_clean to more consistently handle compatible pre-aligned BAMs with different extra contigs in reference genome.
• Fix problem collapsing samples for QC when using RNA-seq variant calling with gatk-haplotype. Thanks to Neill Gibson.
• Integrate ericscript RNA-seq fusion caller. Thanks to Tetiana Khotiainsteva and Vang Le.
• Remove read backed phasing (phasing: gatk) for GATK runs in favor of HaplotypeCaller internal phasing.
• disambiguation: ensure BAM index present for non-split alignments
• Use only end of reads to detect 3' adapters in small RNA-seq pipeline.
• Fix BCBIO_JAVA_HOME to correctly pass custom Java to GATK and Picard runs.
• ChIP-seq: add generation of greylist regions defined as regions of high depth in the input file on a per sample pair basis.
• RNA-seq: STAR now outputs a MAPQ of 60 for uniquely mapped reads instead of 255.
• RNA-seq: Ensure BAM files fed into Cufflinks have 255 as the uniquely mapped MAPQ instead of 60 as output by hisat2/STAR/etc.
• RNA-seq: omit duplicate files from stringtie assembly merging. Thanks to @mmoisse for the bug report.
• Add support for peddy (https://github.com/brentp/peddy) for PED file correspondence/ancestry checking.
• ChIP-seq: pass through encode filtered BAMs to upload directory.
• seq2c: pass through mapping_reads.txt file to directory.

1.0.6 (5 November 2017)

• 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.

1.0.5 (11 Sept 2017)

• Add optional downsampling whole genome BAM files to a high maximum coverage (200 times the average coverage) to avoid slow runtimes in collapsed repeats and poly-ATGC regions. Downsampling happens in parallel with post alignment sorting. Currently turned off by default pending runtime improvements. Configure using maxcov_downsample.
• Separate post alignment recalibration and realignment. Recalibration now occurs multicore to support GATK4 implementation. We generally recommend skipping realignment.
• Provide multicore read trimming and streaming bgzip fastq output with atropos, replacing cutadapt as the default trimmer.
• hg38 runs do not run bwakit's bwa-postproc.js cleanup scripts unless HLA calling needed. Avoids slowdowns using this postprocessing script when running bwa with multiple cores.
• Tumor-only prioritization uses vcfanno output instead of GEMINI, allowing use without needing to build a full GEMINI database.
• Use samtools multicore indexing, replacing sambamba multicore index.
• Replace components of pipeline using single core sambamba view -c with parallel samtools equivalents.
• Replace sambamba depth coverage calculations with mosdepth to improve speed and parallelization.
• Multicore base quality score recalibration with GATK4 and Sentieon.
• GATK4: add support for gVCF based joint calling.
• GATK4: fix option usage for gVCF creation with HaplotypeCaller
• Allow overriding Java used in bcbio with BCBIO_JAVA_HOME
• Do not split individual sample VCFs during pooled batch calling. This previously happened only for small batches with less than 5 samples, now we avoid it entirely and let users do downstream sample extraction.
• Update OptiType HLA calling to use multicore CBC solver, also avoiding GLPK issues.
• Additional approach to retrieving cluster IP addresses for IPython and logging, using the fully qualified domain name.
• Add archive: [cram-lossless] to do CRAM archiving of outputs without quality score compression. Thanks to Alison Meynert.
• Add tools_off: [lumpy-genotype] option to skip Lumpy genotyping.
• CWL/WDL: use single file tarballs for complex collections of files like aligner, RTG and snpEff indices.
• GC bias correction is now the default for Salmon read-based quantification. See https://github.com/salmonteam/SalmonBlogResponse/blob/master/SalmonBlogResponse.md for the reasoning behind this change.
• Add kallisto support for non single-cell RNA-seq experiments.
• Salmon can now be run alongside other RNA-seq quantifiers.
• Cufflinks and Stringtie can be run alongside each other as RNA-seq quantifiers.
• Check BED input files for coordinates off the ends of contigs.

1.0.4 (9 July 2017)

• Initial support for GATK4 variant calling with HaplotypeCaller and MuTect2. Requires tools_on: [gatk4] https://github.com/bcbio/bcbio_validations/tree/master/gatk4
• Enable adapter trimming for variant calling pipeline.
• Provide trim_ends command to quickly do defined end trimming as part of variant calling fastq preparation.
• Support duplex UMIs, present as embedded barcodes on read 1 and read 2.
• Sort region based analyses like variant calling by interval size. Ensures longest intervals run first avoiding delay at end of sample processing.
• Ensure FreeBayes dbSNP and GATK annotations passed into final file. Thanks to @semal.
• Use new Ensembl vep (variant effect predictor) with updated annotations. Thanks to Matthias De Smet.
• Accept files from HTTP/FTP as input
• CWL: use json input files for passing inputs instead of flattened command line arguments. Improves compatibility with multiple runners.
• Allow subsetting a pre-aligned BAM to only standard chromosomes, removing non chr1-22,X,Y for human. This allows runs of pre-aligned data with different extra chromosomes than the bcbio reference builds. Thanks to Oliver Hofmann.
• Improved support for pre-aligned BAMs by using contigs in BAM file for coverage calculations.
• Avoid grabix race conditions with multiple identical input files. Thanks to Andrey Tovchigrechko.
• Remove usage of lxml for qsignature and qualimap to avoid icu library errors.
• CNVkit: merge adjacent calls with identical copy numbers
• Add support for triple-barcoded cellular barcodes.
• Add support for Illumina's SureCell single-cell RNA-seq.

1.0.3 (7 May 2017)

• Allow installs to pull a specific git hash or tag revision of bcbio codebase.
• Fix FreeBayes somatic and multi-sample calling order to be consistent between chromosome region runs. Thanks to Ho Danliang.
• Fix structural variant output upload for complex batching cases. Correctly handle shared normals and other multi-batch by naming outputs using batches. Thanks to Sven-Eric Schelhorn.
• Move to samtools/bcftools/htslib 1.4. Provides parallel bgzip, removing need for pbgzip and improved concatenation speed for region split VCF files.
• Improve Lumpy prioritization speeds by adjusting location of breakend genotyping.
• UMI consensus: reduce runtimes to ~2/3 of previous avoiding unnecessary compression and file IO.
• UMI consensus: pass along metrics about consensus read generation as BAM tags in final file (cD = depth, cE = error rate)
• Support DNApi for de novo adapter detection in small RNA pipeline
• Several updates to the VarScan support: honor options specified in the resource config section; honor min_allele_frac option and set --strand-filter flag in the single-sample case; general cleanups. Thanks to Christian Brueffer.
• Update validation plots to support matplotlib 2.0.
• Enable mixed list/string inputs to germline calling. Thanks to Luca Beltrame.
• Fix qsignature outfile parsing. Thanks to Oliver Hofmann.
• Allow structural variant validations with VCF truth sets. Enables more flexible comparisons without size and event binning.
• Provide seq2c VCF output and enable validation of calls.
• Allow specification of seq2c options through resources. Thanks to Sally Luke and Marisa Cunha.
• Avoid using R_LIBS settings for R runs to limit incompatibilities with externally installed R packages.
• Provide absolute paths for relative paths to files in algorithm list inputs. Thanks to Matthias De Smet.
• Switch to Salmon from Sailfish as default alignment-free RNA-seq quantification algorithm.
• Add sailfish as a valid option for expression_caller.
• Fix chimeric alignment output option for STAR.
• Remove deprecated tidy counts for Sailfish/Salmon.
• Allow more possible empty/skip inputs in variantcaller and svcaller: None, null and empty lists
• Move DEXSeq to be an opt-in expression caller by default.
• Speed up combination of counts/RPKM/FPKM/TPM of samples into a single table by 10x.

1.0.2 (7 March 2017)

• Fix FreeBayes paired somatic calling by generalizing support for finding non-ordered tumor/normal placement in VCF.
• Re-add checks for pre-bgzipped fastq inputs to alignment preparation thanks to a fix for grabix to handle Illumina bgzip outputs.
• Provide DNA damage annotation for low frequency sequencing errors in somatic samples. Use tools_on: [damage_filter]
• Add viral detection for variant calling DNA-seq cancer samples. Uses virus sequences from TCGA GDC distribution and provides simple counts of unmapped reads against viral sequences in MultiQC report.
• Improve lumpy structural variant runs from pre-aligned BAM files, using extract_sv_reads to avoid need to resort input files. Thanks to Neill Gibson.
• Move VCF files from SV prioritization to final upload directory. Thanks to Miika Ahdesmaki.
• Provide whole genome coverage plots with goleft indexcov. Thanks to Brent Pedersen.
• Speed up post-alignment callability calculations by using default parameters to goleft depth. Thanks to Brent Pedersen.
• Allow custom vcfanno configuration files for variant annotation and GEMINI database creation, using vcfanno configuration parameter. Optionally allows use of vcfanno without GEMINI database creation.
• Always use specified cores for analysis re-runs in local multicore mode. Avoids confusing core behavior with checkpoints on re-starts of analysis in a previous work directory.
• Upload of pipeline results to iRODS. Thanks to Matthias De Smet.
• Add duplicate removal to post-FreeBayes processing. Thanks to Neill Gibson.
• Support latest svtyper (0.1.1) for lumpy to provide speed improvements. Will default to 0.1.1 at next release. (use bcbio_conda install -c bioconda svtyper=0.1.1 to test in development)
• Work towards supporting a Python 3 compatible bcbio codebase. Thanks to Michael Crusoe.
• Reduce VarDict maximum BED region sizes for better memory usage. Thanks to Nikolai Karulin, Oliver Hofmann, Miika Ahdesmaki and Zhongwu Lai.
• Require tools_on: [lumpy_usecnv] to pre-run CNVkit as input to Lumpy, allowing Lumpy and CNVkit to run in parallel otherwise.
• Avoid issues with CNVkit bin size estimates for normal associated with multiple tumors. Thanks to Ho Danliang.
• Fix double uploading of fast RNA-seq quantification.
• Output single-cell RNA-seq counts in annotated MatrixMarket format.

1.0.1 (17 January 2017)

• Fix bug in 1.0.0 release with parallel calculations on whole genome samples. The release version only parallelizes by chromosome instead of callable regions, resulting in less parallelism. Thanks to Sven-Eric Schelhorn and Neill Gibson.
• Generalize use of working directories to support runs on S3 mounted filesystems. Ensures all work takes place inside transactional directories. Thanks to Tetiana Khotiainsteva and Sven-Eric Schelhorn.
• Provide separate germline calling for somatic tumor/normal pairs. Supplements somatic calls with standard germline calls on normal samples, including ensemble and SV calling.
• Support creating GEMINI databases with new generic mechanism using vcfanno/vcf2db. This allows creation of GEMINI output for any organism. Adds support for hg38 with annotations from dbSNP, Clinvar, ExAC and ESP.
• Support FreeBayes 1.1.0 for improved memory usage and 3-4x speedup. Will default to 1.1.0 at next release. Validation work: https://github.com/bcbio/bcbio.github.io/blob/master/_posts/2016-11-21-giab-hg38-freebayes.md
• Rework quality control for speed and output directory consistency. Avoid re-duplicating calculations and put all output in qc directory to make re-runs easier. Thanks to Vlad Saveliev.
• Fixes for Seq2C concurrency problems when preparing BED files. Thanks to Vlad Saveliev.
• Update WHAM structural variant caller to support the latest release.
• Update delly structural variant caller to support the latest release.
• Improve dbSNP annotation speeds for adding rs IDs to VarDict output. Thanks to Ben Liesfeld.
• Support for VEP 87 with additional plugins and generalization of fields. Thanks to Matthias De Smet.
• Deprecate clinical_reporting parameter and introduce new effects_transcripts parameter than enables more control over variant effects prediction. Enable HGVS by default for human projects and separates from transcript selection.
• For lumpy runs that use samblaster, use samtools sort instead of sambamba sort. Avoids segfault issues with samblaster. Thanks to Oliver Hofmann.
• Pre-install capture region BED files and enable short hand specification in sample configuration.
• Use vt normalize as part of GEMINI decomposition to clean up complex multiallelic variants. Thanks to Sergey Naumenko.
• Testing suite cleanup. Move to py.test and separate integration and unit tests. Thanks to Tetiana Khotiainsteva.
• Fix issue with cutadapt hanging on gzipped input. Thanks to Stephen Turner.
• Updated cutadapt to use single-pass trimming for paired-end files, improving performance and hitting the disk less.
• Added support for cellular barcode error correction with single-cell RNA-seq via the cellular_barcode_correction parameter. This corrects edit distances up to the set value, defaults to 1.
• Add support for sample-based demultiplexing of single-cell RNA-seq runs.
• Move single-cell RNA-seq results to the upload directory.
• Make positional UMI default to off for single-cell RNA-seq.
• Add support for the Klein lab v3 version of the inDrop protocol.

1.0.0 (20 November 2016)

• Default to no calling if variantcaller not specified, instead of old GATK UnifiedGenotyper default.
• Use samtools depth instead of bedtools genomecov for depth calculations, and calculate high depth regions during initial depth calculations. Improves speed by more than 6x. Thanks to Brent Pedersen.
• Adjust de-duplication strategy to use bamsormadup from biobambam2 for most cases and samblaster when split and discordant reads needed for SV calling with lumpy.
• Fix handling of fresh installs with GATK 3.6 only included. Correctly handles versioning from bioconda and lack of specifically defined jar directory.
• Unset JAVA_HOME when running gatk-framework and GATK > 3.6, forcing use of bcbio installed Java 8. Thanks to Brad Wubbenhorst.
• Fix bug when running realignment without recalibration in GATK 3.6. Thanks to Pär Larsson.
• Get from GEO server, GSM FASTQ samples using bcbio_prepare_samples.py script
• Add seqcluster stats to QC folder
• Allow manual specification of total memory and core usage for machines in resources. Thanks to Juan Caballero.
• Allow PED based gender specifications (1=male, 2=female). Thanks to Brent Pedersen.
• Annotate validation variants with genome context from GA4GH and other sources for interpreting true/false positives/negatives.
• Limit GATK cores used for GenotypeGVCFs to avoid excessive memory usage.
• VQSR: allow forcing GATK to try VQSR with tools_on. Generate VQSR plots. Thanks to Zhengqiu Cai.
• Support ATAC-seq for chipseq pipeline.
• Remove duplicates after alignment for chipseq pipeline.
• Support for bzip2 input files during variant calling. Thanks to Paulo Silva.
• Allow non-positional UMI Rapmap quantified single-cell RNA-seq.
• Re-enable save_diskspace option to reduce disk usage during alignment preparation and split alignments.
• Offload fixing the unmapped Tophat file to tophat-recondition. Thanks to Christian Brueffer.
• Add support for vcfanno (https://github.com/brentp/vcfanno) to annotate VCFs with other VCFs/BED files.
• Mark possible RNA-edits for GRCh37/hg19 using RADAR coordinates. Thanks to Sergey Naumenko for the suggestion.
• Add local_controller option to run the controller alongside the main bcbio process. Thanks to Brent Pederson and Sven-Eric Schelhorn.

0.9.9 (18 July 2016)

• Change defaults for recalibration and realignment to False. These have been the recommended settings (http://bit.ly/bcbio-minimal) and no realignment now matches Broad recommendations.
• Use conda installed Java instead of requiring external installation for most tools.
• Support GATK 3.6 with Java 8 installed as part of anaconda. Older GATK versions for calling and recalibration/realignment require external Java 7.
• Re-organization of variants stats using bcftools and cleaning gemini queries to get individual samples metrics.
• Quality control back end revamped to support better parallelization and pluggability of new QC metrics.
• Support CNV calling with Seq2C for exome, targeted or amplicon experiments. Thanks to Vlad Saveliev.
• Add fixrg target to bam_clean to accept BAM inputs with correct sorting and reads but that need an updated read group.
• More robust file transactions across network filesystems, avoiding failures from partially transferred files. Thanks to Sven-Eric Schelhorn.
• Improved checking of BAM files during merge steps. Thanks to Sven-Eric Schelhorn.
• Add SAMPLE and PEDIGREE tags to tumor/normal VCF outputs to enable easier post-analysis parsing of results.
• Add single point for annotation following variant calling to improve pluggability of new annotation types.
• Add support for running germline and somatic calling with Sentieon callers (https://peerj.com/preprints/1672/). Requires license from Sentieon.
• Fix fusion calling using Tophat2. Thanks to @csardas for raising the issue.
• Add support for kallisto quantification of single-cell RNA-seq data.
• Add transcriptome_fasta option to single-cell RNA-seq. This allows the user to provide a transcriptome FASTA file to quantitate against rather than use the bcbio provided annotation.
• Fix naming of vardict RNA-seq variant calls. Thanks to @csardas.

0.9.8 (20 May 2016)

• Correctly install all datatargets on new installation. Previously we'd skipped installing default additional data unless specified.
• Use yamllint to find wrong syntaxes in the YAML file that are ignored by pyyaml package and can affect the analysis.
• Improve choosing split regions for batch analysis to use the unionized intersection of non-callable regions. This enables better use of batches with different callable regions. Thanks to Neill Gibson.
• Fix HLA typing issues and handle HLA typing on split alignments. Thanks to Miika Ahdesmaki.
• Set align_split_size automatically based on input file sizes, trying to provide reasonable splits and avoid too many splits for large files.
• Fix high depth identification for whole genome runs, correctly calculating it when also inferring coverage estimations. Thanks to Neill Gibson.
• Do not remove duplicates for GATK variant calling when mark_duplicates is False or running amplicon sequencing.
• Fix installation of mutect jar via toolplus when mutect not previously present in configuration.
• Platypus: revert filtering back to defaults after additional cross-validation: http://i.imgur.com/szSo5M6.png
• Enable gVCF output with tools_on: [gvcf] for users who need gVCF output for downstream analyses.
• Avoid downscaling memory when recalibrating/realigning with GATK, since we should not longer need to work around Java issues. Thanks to Luca Beltrame.
• Do not use samblaster on genomes with greater than 32768 contigs, the samblaster maximum. Thanks to morten (@mattingsdal).
• Move to samtools for output CRAM support, using bamUtils for 8-bin compression of read quality scores.
• Remove merge_bamprep option and always merge realigned BAM files if run.
• Correctly clean up additional problem characters in sample descriptions that can confuse shell commands.

0.9.7 (29 March 2016)

• Use MultiQC (github.com/ewels/MultiQC) as main package to process all QC metrics.
• New install procedure for data: --datatarget allows installation of sub-sets of supplemental data for smaller installs for small RNA only analysis. Also provides a consistent framework for installing larger data types.
• VEP data no longer installed by default. Requires --datatarget vep
• During install, --toolplus only used for third party tools like GATK and MuTect and not data installation, which moved to --datatarget
• Provide data_versions.csv in output folder that has versions of reference data used in the analysis.
• Use sample description for BAM read group IDs, instead of lane index. This allows remixing of samples after processing without potential collisions. Thanks to Neill Gibson.
• Use sample description for file names instead of lane/flowcall information. Makes re-runs more stable when using template and files easier to interpret. Back compatible with re-runs of old work directories.
• Finalize support for MuTect2 with validation against the DREAM synthetic 4 dataset (http://imgur.com/CLqJlNF). Thanks to Alessandro (@apastore).
• Do not bgzip inputs when they are already gzipped and do not require parallelization or format conversion. Thanks to Miika Ahdesmaki.
• Use new snpEff annotations (ANN) instead of older approach (EFF). The new annotations are more interoperable and supported by GEMINI.
• Lazy import of matplotlib libraries to avoid slow startup times.
• Only apply ploidyfix to all female batches to remove Y chromosome. Avoids confusion with file produced in other cases without any changes.
• Improvement to bcbio CWL integration: support parallel alignment and variant calling.
• Support for Salmon and RapMap added.
• FastRNA-seq pipeline implemented that does nothing but run Salmon with no QC.
• Singlecell RNA-seq pipeline implemented that uses https://github.com/vals/umis to handle the UMI and cellular barcode, aligns with RapMap and quantitates by counting, scaling ambiguous reads by the number of transcripts they could have come from.
• Migrate bowtie and bowtie2 to handle split input alignments, bgzipped inputs, and produce sorted, de-duplicated BAM files. This allows use in additional standard pipelines. Thanks to Luca Beltrame.
• Switch final upload directories for salmon and sailfish results to be of the form samplename/salmon instead of samplename/salmon/samplename.

0.9.6 (12 February 2016)

• Installation uses conda packages from bioconda for Python dependencies and third party tools.
• Add macs2 to chipseq pipeline.
• Add germline output files for somatic calling pipelines. The standard variant calls identify somatic mutations different from a normal, while the germline has pre-existing mutations which might contribute to cancer development.
• Use parallel bgzip for preparation of input fastq files for parallelization and alignment. Thanks to Guillermo Carrasco.
• Avoid extacting individual sample calls from pooled variant call runs for samples with more than 5 individuals in a batch. Avoids slow extraction run times. Thanks to Neill Gibson.
• Add explicit check for BED file mismatches with reference genome.
• During validation, report truth counts relative to initial truth set representation and pick best metric for plotting ROC scores.
• Remove --sudo flag from installer. bcbio requires install into a directory structure with user permissions.
• Add ability to tweak fastq preparation for alignment splitting so we can explore alternative approaches to bgzip and grabix index.
• Re-enable stringtie as an expression caller.
• Allow stringtie as a transcript assembler.
• Replace the assemble_transcriptome option with transcript_assembler, which accepts a list of assemblers to run. The output of all the assemblers is merged at the end with Cuffmerge.
• Move Picard to use conda installed picard single executable instead of custom installed java directory of jars.
• Add library type option to Cufflinks assembly. Thanks to Konstantin (@dezzan).
• Tag variants decomposed with vcfallelicprimitives. Thanks to Neill Gibson.
• Fix Platypus problem where we weren't correctly specifying BED regions since latest update skips over files not ending with".txt" or ".bed".

0.9.5 (12 December 2015)

• Add miRDeep2 to small RNA-seq analysis and quantify the novel miRNAs for all samples.
• Enable calling of HLA alleles with human build 38 (hg38). Turn on with the hlacaller option.
• Structural variant prioritization with BED files of known biologically important regions. Extracts SV calls in these regions and produces a tab delimited high level summary. Use the svprioritize option to enable.
• Add tRNA count and figures by tdrmapper for srna-seq pipeline.
• Avoid running callability checks on smaller chromosomes less than 1 million basepairs. Saves computation and disk IO on alt and support regions we don't split on.
• Enable nested batch specifications, allowing samples in partially overlapping batches.
• Speed improvements for Lumpy genotyping. Move to latest svtyper and avoid genotyping breakends.
• Allow use of VEP annotations on non-human analyses.
• Filter VarDict calls with poor mapping quality support (-Q 10) which trigger low frequency false positives.
• Remove ENCODE blacklist regions when calling with VarDict and FreeBayes on whole genomes. Avoids long run times due to collapsed repeats near centromeres.
• Update VarScan to 2.4.0 and rework support to allow piping between mpileup and VarScan to avoid filesystem IO.
• Annotate ensemble calls with information about supporting callers. Thanks to Pär Larsson and Son Pham.
• Move eXpress to expression_caller instead of being run by default.
• rRNA calculation uses the count file instead of using counts from GATK.
• Merge STAR fusion calls back into the BAM file. Thanks to Miika Ahdesmaki.
• Added preliminary support for the hisat2 aligner.
• Swapped STAR indexing to use on the fly splice junction indexing.
• Slightly inceased default DEXseq memory requirements in bcbio_system.yaml.
• Add support for RNA-seq for hg38 and hg38-noalt
• Make Sailfish the default for non-count based expression estimation. Produces isoform-level (combined.isoform.sf.tpm) and gene-level (combined.gene.sf.tpm) TPM expression estimation.
• Move Cufflinks to be off by default for expression estimation (turn on via expression_callers if needed).
• Add STAR fusion gene parameters suggested by @felixschlesinger.
• Add disambiguation to Sailfish by creating a master FASTA file of all transcripts from all organisms, quantitating each and separating out the organism-specific transcripts after.
• Add VarDict support for RNA-seq variant calling. Thanks to Miika Ahdesmaki and Sven-Eric Schelhorn.

0.9.4 (14 October 2015)

• Ensure genome data sort order is identical to BED files when annotating structural variant calls. Thanks To Miika Ahdesmaki.
• Improve low frequency calling for VarDict using vaidation against DREAM synthetic dataset 4.
• Install truth sets for germline and cancer calling automatically as part of bcbio and make it easy to include them in the configuration files for validation.
• Avoid need to set LD_LIBRARY_PATH and PERL5LIB on installations.
• Update Scalpel to latest version (0.5.1) and improve sensitivity for low frequency indels: http://imgur.com/a/7Dzd3
• Drop coverage_depth_max for downsampling, which no longer works in GATK 3.4. The option wasn't supported by other callers so was more confusing than useful.
• Fix missing BAM index when running with align: false. Thanks to Stephan Pabinger and Severine Catreux.
• Annotate structural variant files with snpEff. Initial steps towards summarized structural variant reporting.
• Add ability to specify platform unit (PU) and library (LB) in BAM header. Thanks to Brad Wubbenhorst.
• Update gatk-framework to 3.4-46 to avoid errors dealing with new gVCF output.
• Set java.io.tmpdir to avoid filling up global temporary space with snpEff. Thanks to Oliver Hofmann.
• Speed up transcriptome-only processing. Thanks to Sven-Eric Schelhorn.
• Add bamtools output to RNA-seq quality metrics. Thanks to Sven-Eric Schelhorn.
• Expand input quality format detection to detect full range of possible Sanger values.

0.9.3 (27 September 2015)

• Fix bug when using tumors with multiple normals and no CNV calling. Additional tumor sample would get lost due to lack of early (CNV-based) calling. Thanks to Miika Ahdesmaki.
• Include R and Rscript in the installation with conda packages and use for installing and running R-based tools. Avoids issues with alternative R versions and need for a separate installation.
• Fix bug when using CNVkit on disambiguated inputs. Thanks to Miika Ahdesmaki.
• Re-work structural variant infrastructure to provide plug-in parallel ensemble calling, removing the previous overlap-based ensemble calls. Currently supports MetaSV for ensemble calls. Also re-works validation to not rely on ensemble-overlap calls.
• Default to using Real Time Genomics vcfeval (https://github.com/RealTimeGenomics/rtg-tools) for validation instead of bcbio.variation. Improves speed and resolution of closely spaced variants. The old funtionality is still available with validate_method: bcbio.variation.
• Correctly apply BQSR when using recalibration with PrintReads by using GATK full instead of the open source GATK framework which silently ignores BQSR option. Thanks to Severine Catreux.
• Require larger blocks (250bp, moved from 100bp) to find regions for splitting analysis to avoid too tight splitting around small homozygous deletions.
• Adjust mapping quality (MQ) filter for GATK SNP hard filters to improve sensitivity http://imgur.com/a/oHRVB
• Ensure memory specification passed to sambamba and samtools sort during disambiguation and RNA-seq. Thanks to Sven-Eric Schelhorn.
• Fix compatbility with bedtools groupby in v2.25.0, which needs short parameters instead of long parameter names.
• Allow turning off variant quality score recalibration with tools_off: [vqsr]
• Generalize group size for batching gVCFs prior to joint calling with joint_group_size. Thanks to Severine Catreux.
• Support GEMINI 0.17.0, which does not have a --no-bcolz option since that is the default.
• Remove test_run parameter since it was poorly supported and not used much.
• Fix issue with featureCounts sorting not working in parallel by pre-sorting and filtering the BAM file.
• Unified stock coverage and experimental coverage reporting.
• Deprecated report and coverage_experimental as algorithm keys.

0.9.2 (1 September 2015)

• Support IPython 4.0 with ipyparallel
• Fix bug in writing BAM and VCF indexes to final directory. Correctly add indexes as bam.bai and vcf.gz.tbi.
• Fix bug in queryname sorting on split files for feeding into diambiguation. Ensure proper sorting with explicity sambamba sort. Thanks to Sven-Eric Schelhorn.
• Ensure extra FreeBayes alleles get removed prior to vcfallelicprimatives, avoiding leaving incorrect genotype allele fields. Thanks to Michael Schroeder.
• Split CNVkit processing into individual components, enabling better parallelization and control over parameters.
• Genotype Lumpy structural variant calls with SVtyper.
• Initial support for small RNA pipeline. Thanks to Lorena Pantano.
• Support for MetaSV to prepare combined structural variant calls.
• Add smallRNA-seq pipeline
• Test automatic report for variants calling and standard pipeline.
• Allow Cufflinks to be turned off via tools_off.

0.9.1 (6 August 2015)

• Fix novoalign to work with parallel split alignments. Thanks to Tyler Funnell.
• Move lumpy-sv to latest version which uses lumpyexpress instead of speedseq.
• Support the manta SV caller from Illumina. Validations: http://imgur.com/a/Gajsg
• Remove high depth regions from structural variant calling exclusion file to avoid false positives with lumpy. Thanks to Miika Ahdesmaki.
• Move some structural variant calling, like CNV detection, prior to variant calling. Allows use of CNV calls as inputs for variant detection tools.
• Generalize support for interaction with blob storage and graphing to support alternative cloud providers. Initial support for interacting with Azure. Thanks to Alexandru Coman.
• Remove VarDict call lines where reference and alternative allele are identical.
• Fix assignment issues during prioritization with new GEMINI and sqlite.
• Support updated versions of sambamba, which provide headers for window depth commands.

0.9.0 (20 June 2015)

• GATK 3.4: support HaplotypeCaller by avoiding setting downsampling (-dcov) option by default.
• Single sample structural variant calling: corectly handle multiple variant callers. Thanks to Sven-Eric Schelhorn.
• Make VarDictJava the default caller when vardict specified. vardict-perl is now required to specifically use the Perl version.
• VarDict and VarDictJava: limit regions to 1Mb with overlaps to avoid memory errors. Ignore regions without BED reads which can lead to large genomic sections and memory errors.
• VarDict and VarDictJava: annotate outputs with dbSNP.
• Add tools_on configuration with svplots option. This turns off structural variant plotting by default, which can be time consuming compared to calling.
• Add a --only-metadata argument to template preparation that will only include BAM or fastq files in sample YAML if they are present in the metadata CSV file.
• samblaster: support -M flag in 0.1.22 release
• Fix VEP/GEMINI incompatibility where empty fields are included in VCF output.
• VarDict: restrict maximum region size within a BED file to 2Mb to avoid high memory usage and failures for longer regions.
• Include snpEff effects summary file in output directory when used for effects prediction.

0.8.9 (10 May 2015)

• Upgrade variant effect predictor (VEP) to the latest Ensembl version (79) with support for hg38. The latest VEP has better support for multiple versions but incompatible database naming. This requires an update of tools and data in a two step process. First bcbio_nextgen.py upgrade -u stable --tools (or -u development) then bcbio_nextgen.py upgrade --data.
• Improve de-duplication for split alignments. Do not sort/merge during splits, and instead perform a global merge sort and de-duplication of the final set of reads.
• Initial support for new human genome build (hg38/GRCh38) including alternative alleles. Usage is in place but still requires validation and additional testing.
• Remove alternative alleles from downstream variant calling after using in alignment to avoid issues with chromosome names like HLA*.
• Enable installation of external conda-managed tools. Adds in builds for heterogeneity analysis.
• Clean up preparation process for multi-allelic inputs to GEMINI to avoid needing to split/merge. Thanks to Sven-Eric Schelhorn.

0.8.8 (29 April 2015)

• Automatically calculate coverage_interval based on coverage calculations, avoiding need to set this directly in input configuration.
• Update vt decompose to handle additional multi-allelic adjustments including all format attributes, providing full support for new GEMINI changes. Thanks to Brent Pedersen and Adrian Tan.
• Add default configuration target to bcbio_system.yaml reducing the need to set program specific arguments for everything.
• Ensure resources specified in input YAML get passed to global system configuration for making parallelization decisions. Thanks to Miika Ahdesmaki.
• Run upload process on distributed machines, allowing upload to S3 on AWS to take advantage of machines with multiple cores. Thanks to Lorena Pantano.
• Re-write interactions with external object stores like S3 to be more general and incorporate multiple regions and future support for non-S3 storage.
• Scale local jobs by total memory usage when memory constrains resource usage jinstead of cores. Thanks to Sven-Eric Schelhorn and Lorena Pantano.
• Disambiguation: improve parallelization by disambiguating on split alignment parts prior to merging. Thanks to Sven-Eric Schelhorn.
• Disambiguation: ensure ambiguous and other organism reads are sorted, merged and passed to final upload directory. Thanks to Sven-Eric Schelhorn.
• Fix problem with sambamba name sorting not being compatible with samtools. Thanks to Sven-Eric Schelhorn.
• FreeBayes: update to latest version (0.9.21-7) with validation (http://imgur.com/a/ancGz).
• Allow bz2 files in bcbio_prepare_sample.py script.
• Ensure GEMINI statistics run for project summary file. Thanks to Luca Beltrame.
• Better error checking for booleans in input configuration. Thanks to Daryl Waggott.
• Implement qualimap for RNAseq QC metrics, but not active yet.
• collect statistics graphing capabilities moved from bcbio-nextgen-vm, enabling plotting of resource usage during runs. Thanks to John Morrissey and Lorena Pantano.

0.8.7 (12 March 2015)

• Run snpEff 4.1 in back-compatibility mode to work with GEMINI database loading. Fixes snpEff 4.1/GEMINI effects loading.
• Add PED file to GEMINI database load, containing family, gender and phenotype information from bcbio metadata. Thanks to Luca Beltrame and Roy Ronen.
• Enable specification of input PED files into template creation, extracting family, gender and phenotype information. Any sample rows from PED files get used when creating the GEMINI database.
• Fix preparation of multi-allelic inputs to GEMINI by implementing custom merge of bi-allelic and split multi-allelic. Previous implementation using GATK CombineVariants re-merged some split multi-allelic, losing effects annotations.
• Skip contig order naming checking with bedtools 2.23.0+ to avoid potential issues with complex naming schemes.
• Installation and upgrade: Set pip SSL certificates to point at installed conda SSL package if present. Avoids SSL errors when pip can't find system certificates. Thanks to Andrew Oler.
• Enable support for PBSPro schedulers through ipython-cluster-helper.

0.8.6 (23 February 2015)

• Calculate high depth regions with more than 20x median coverage as targets for filtering in structural variants. Attempts to detect and avoid spurious calls in repetitive regions.
• Support snpEff 4.1, including re-download of snpEff databases on demand if out of sync with older versions.
• Split multi-allelic variants into bi-allelic calls prior to loading into GEMINI, since it only handles bi-allelic inputs. Thanks to Pär Larsson.
• Pass ploidy to GATK HaplotypeCaller, supporting multiple ploidies and correct calling of X/Y/MT chromosomes. Requires GATK 3.3.
• Remove extra 'none' sample when calling tumor-only samples using MuTect. Harmonizes headers with other tumor-only callers and enables tumor-only ensemble calling. Thanks to Miika Ahdesmaki.
• Perform variant prioritization as part of tumor-only calling, using population based frequencies like 1000 genomes and ExAC and presence in known disease causing databases like COSMIC and Clinvar.
• Switch to samtools sort from sambamba sort during alignment streaming. Saves steps in processing and conversions on single sample no deduplication inputs.
• On AWS, download inputs for S3 instead of streaming into fastq preparation to avoid issues with converting BAM to fasta. Thanks to Roy Ronen.
• Provide better defaults for mincores that packs together multiple single IPython processes on a single cluster request -- use core specification from input configuration. Thanks to Miika Ahdesmaki.

0.8.5 (11 January 2015)

• No longer keep INFO fields with vcfallelicprimitves in FreeBayes, Platypus and Scalpel calling to prevent introduction of problematic fields for multi-allelic MNPs.
• Fix batching problem when using coverage and multiple shared batches like a global normal in cancer calling. Thanks to Luca Beltrame.
• Use mincores specification to ipython-cluster-helper to combine single core jobs into a single submission job for better memory shared on resource constrained systems.
• Move disambiguation split work inside parallel framework so download and preparation occurs on worker nodes or inside Docker containers. Enables on demand download of disambiguation genomes.
• Ensure population databases created when some inputs do not have variant calls.
• Switch to seaborn as matplotlib wrapper, from prettplotlib.
• Fixes for ensemble structural variant calling on single samples.
• Fixes for mixing joint and pooled calling in a single configuration file.
• Support for qSNP for tumor-normal calling.
• Add eXpress to RNA-seq pipeline.
• Add transcriptome-only mapping with STAR, bowtie2 or bwa.
• Change logging time stamps to be UTC and set explicitly as ISO 8601 compliant output. Improves benchmarking analysis and comparability across runs.
• Add support for RNA-seq variant calling with HaplotypeCaller
• Fix parallelization of DEXSeq.

0.8.4 (29 November 2014)

• Improvements in VarDict calling on somatic samples.
• Fix compatibility issue with bedtools 2.22.0 when calculating genome coverage.
• Fix joint calling upload to avoid redundant inclusion of full VCF file in individual sample directories.
• Fixes for inclusion of GATK jars inside Docker contains when running distributed jobs.
• Enable generation of STAR indexes on demand to handle running STAR on AWS instances.
• Re-organize code to prepare samples and reference genomes so it runs inside distributed processing components. This isolates process to Docker containers on AWS and also enables complex operations like preparing reference genomes on demand.

0.8.3 (19 November 2014)

• Improve tumor/normal calling with FreeBayes, MuTect, VarDict and VarScan by validating against DREAM synthetic 3 data.
• Validate ensemble based calling for somatic analysis using multiple callers.
• Improve ability to run on Amazon AWS, including up to date interaction with files originally stored in S3 and transfer to S3 on completion with encryption.
• Avoid race conditions during bedprep work on samples with shared input BED files. These are now processed sequentially on a single machine to avoid conflicts. Thanks to Justin Johnson.
• Add data checks and improved flexibility when specifying joint callers. Thanks to Luca Beltrame.
• Default to a reduced number of split regions (nomap_split_targets defaults to 200 instead of 2000) to avoid controller memory issues with large sample sizes.
• Avoid re-calculating depth metrics when running post variant calling annotation with GATK to provide accurate metrics on high depth samples. Thanks to Miika Ahdesmaki.
• Consistently keep annotations and genotype information for split MNPs from vcfallelicprimitives. Thanks to Pär Larsson.
• Enable VQSR for large batches of exome samples (50 or more together) to coincide with joint calling availability for large populations.
• Support retrieval of GATK and MuTect jars from S3 to enable integration with bcbio inside Docker.
• Bump pybedtools version to avoid potential open file handle issues. Thanks to Ryan Dale.
• Move to bgzipped and indexes human_ancestor.fa for LOFTEE to support access with new samtools that no longer uses razip.

0.8.2 (September 17, 2014)

• Fix bug in creating shared regions for analysis when using a single sample in multiple batches: for instance, when using a single normal sample for multiple tumors. Thanks to Miika Ahdesmaki.
• Unify approach to creating temporary directories. Allows specification of a global temporary directory in resources: tmp: used for all transactions. This enables full use of local temporary space during processing, with results transferred to the shared filesystem on completion.
• Fix issues with concatenating files that fail to work with GATK's CatVariants. Fall back to bcftools concat which correctly handles problem headers and overlapping segments.
• Enable flexible specification of indelcaller for variantcaller targets that do not have integrated indel methods. Thanks to Miika Ahdesmaki.
• Move to samtools 1.0 release. Update samtools variant calling to support new multiallelic approach.
• Improve Platypus integration: correctly pass multiple BAM files, make use of assembler, split MNPs, and correctly restrict to variant regions.
• Be more aggressive with system memory usage to try and make better use of available resources. The hope is to take advantage of Java memory fixes that previously forced us to be conservative.

0.8.1 (August 29, 2014)

• Support joint recalling with GATK HapolotypeCaller, FreeBayes and Platypus. The jointcaller configuration variable enables calling concurrently in large populations by independently calling on samples them combining into a final combined callset with no-call/reference calls at any position called independently.
• Add qsignature tool to standard and variant analyses, which helps identify sample swaps. Add mixup_check configuration variant to enable.
• Fix issue with merging GATK produced VCF files with vcfcat by swapping to GATK's CatVariants. Thanks to Matt De Both.
• Initial support for ensemble calling on cancer tumor/normal calling. Now available for initial validation work. Thanks to Miika Ahdesmaki.
• Enable structural variant analyses on shared batches (two tumors with same normal). Thanks to Miika Ahdesmaki.
• Avoid Java out of memory errors for large numbers of running processes by avoiding Parallel GC collction. Thanks to Justin Johnson and Miika Ahdesmaki.
• Enable streaming S3 input to RNA-seq and variant processing. BAM and fastq inputs can stream directly into alignment and trimming steps.
• Speed improvements for re-running samples with large numbers of samples or regions.
• Improved cluster cleanup by providing better error handling and removal of controllers and engines in additional failure cases.
• Support variant calling for organisms without dbSNP files. Thanks to Mark Rose.
• Support the SNAP aligner, which provides improved speed on systems with larger amount of memory (64Gb for human genome alignment).
• Support the Platypus haplotype based variant caller for germline samples with both batched and joint calling.
• Fix GATK version detection when _JAVA_OPTIONS specified. Thanks to Miika Ahdesmaki.
• Use msgpack for ipython serialization to reduce message sizes and IPython controller memory instead of homemade json/zlib approach.

0.8.0 (July 28, 2014)

• Change defaults for installation: do not use sudo default and require --sudo flag for installing system packages. No longer includes default genomes or aligners to enable more minimal installations. Users install genomes by specifically enumerating them on the command line.
• Add support for Ensembl variant effects predictor (VEP). Enables annotation of variants with dbNSFP and LOFTEE. Thanks to Daniel MacArthur for VEP suggestion.
• Support CADD annotations through new GEMINI database creation support.
• Rework parallelization during variant calling to enable additional multicore parallelization for effects prediction with VEP and backfilling/squaring off with bcbio-variation-recall.
• Rework calculation of callable regions to use bedtools/pybedtools thanks to groupby tricks from Aaron Quinlan. Improves speed and memory usage for coverage calculations. Use local temporary directories for pybedtools to avoid filling global temporary space.
• Improve parallel region generation to avoid large numbers of segments on organisms with many chromosomes.
• Initial support for tumor normal calling with VarDict. Thanks to Miika Ahdesmaki and Zhongwu Lai.
• Provide optional support for compressing messages on large IPython jobs to reduce memory usage. Enable by adding compress_msg to alogrithm section of bcbio_system.yaml. There will be additional testing in future releases before making the default, and this may be replaced by new methods like transit (https://github.com/cognitect/transit-python).
• Add de-duplication support back for pre-aligned input files. Thanks to Severine Catreux.
• Generalize SGE support to handle additional system setups. Thanks to Karl Gutwin.
• Add reference guided transcriptome assembly with Cufflinks along with functions to classify novel transcripts as protein coding or not as well as generally clean the Cufflinks assembly of low quality transcripts.
• Developer: provide datadict.py with encapsulation functions for looking up and setting items in the data dictionary.
• Unit tests fixed. Unit test data moved to external repository: https://github.com/roryk/bcbio-nextgen-test-data
• Add exon-level counting with DEXseq.
• Bugfix: Fix for Tophat setting the PI flag as inner-distance-size and not insert size.
• Added kraken support for contamination detection (@lpatano): http://ccb.jhu.edu/software/kraken/
• Isoform-level FPKM combined output file generated (@klrl262)
• Use shared conda repository for tricky to install Python packages: https://github.com/chapmanb/bcbio-conda
• Added initial chanjo integration for coverage calculation (@kern3020): https://github.com/robinandeer/chanjo
• Initial support for automated evaluation of structural variant calling.
• Bugfix: set library-type properly for Cufflinks runs.
• Added genome_setup.py a script to prepare your own genome and rnaseq files.

0.7.9 (May 19, 2014)

• Redo Illumina sequencer integration to be up to date with current code base. Uses external bcl2fastq demultiplexing and new bcbio integrated analysis server. Provide documentation on setting up automated infrastructure.
• Perform de-duplication of BAM files as part of streaming alignment process using samblaster or biobambam's bammarkduplicates. Removes need for secondary split of files and BAM preparation unless recalibration and realignment needed. Enables pre-processing of input files for structural variant detection.
• Rework batched regional analysis in variant calling to remove custom cases and simplify structure. Filtering now happens explicitly on the combined batch file. This is functionally equivalent to previous filters but now the workflow is clearer. Avoids special cases for tumor/normal inputs.
• Perform regional splitting of samples grouped by batch instead of globally, enabling multiple organisms and experiments within a single input sample YAML.
• Add temporary directory usage to enable use of local high speed scratch disk on setups with large enough global temporary storage.
• Update FreeBayes to latest version and provide improved filtering for high depth artifacts.
• Update VQSR support for GATK to be up to date with latest best practices. Re-organize GATK and filtering to be more modular to help with transition to GATK 3.x gVCF approaches.
• Support CRAM files as input to pipeline, including retrieval of reads from defined sequence regions.
• Support export of alignment data as CRAM instead of BAM for space storage and long term archiving.
• Provide configuration option, remove_lcr, to filter out variants in low complexity regions.
• Improve Galaxy upload for LIMS supports: enable upload of FastQC as PDF reports with wkhtmltopdf installed. Provide tabular summaries of mapped reads.
• Improve checks for pre-aligned BAMs: ensure correct sample names and provide more context on errors around mismatching reference genomes.
• GATK HaplotypeCaller: ensure genotype depth annotation with DepthPerSampleHC annotation. Enable GATK 3.1 hardware specific optimizations.
• Use bgzipped VCFs for dbSNP, Cosmic and other resources to save disk space. Upgrade to Cosmic v68.
• Avoid VCF concatenation errors when first input file is empty. Thanks to Jiantao Shi.
• Added preliminary support for oncofuse for calling gene fusion events. Thanks to @tanglingfung.

0.7.8 (March 21, 2014)

• Add a check for mis-specified FASTQ format in the sample YAML file. Thanks to Alla Bushoy.
• Updated RNA-seq integration tests to have more specific tags (singleend, Tophat, STAR, explant).
• Fix contig ordering after Tophat alignment which was preventing GATK-based tools from running.
• Allow calculation of RPKM on more deeply sampled genes by setting --max-bundle-frags to 2,000,000. Thanks to Miika Ahdesmaki.
• Provide cleaner installation process for non-distributable tools like GATK. The --tooplus argument now handles jars from the GATK site or Appistry and correctly updates manifest version information.
• Use bgzipped/tabix indexed variant files throughout pipeline instead of raw uncompressed VCFs. Reduces space requirements and enables parallelization on non-shared filesystems or temporary space by avoiding transferring uncompressed outputs.
• Reduce memory usage during post-alignment BAM preparation steps (PrintReads downsampling, deduplication and realignment prep) to avoid reaching memory cap on limited systems like SLURM. Do not include for IndelRealigner which needs memory in high depth regions.
• Provide explicit targets for coverage depth (coverage_depth_max and coverage_depth_min) instead of coverage_depth enumeration. Provide downsampling of reads to max depth during post-alignment preparation to avoid repetitive centromere regions with high depth.
• Ensure read group information correctly supplied with bwa aln. Thanks to Miika Ahdesmaki.
• Fix bug in retrieval of snpEff databases on install. Thanks to Matan Hofree.
• Fix bug in normal BAM preparation for tumor/normal variant calling. Thanks to Miika Ahdesmaki.
• General removal of GATK for variant manipulation functionality to help focus on support for upcoming GATK 3.0. Use bcftools for splitting of variants into SNPs and indels instead of GATK. Use vcflib's vcfintersection to combine SNPs and indels instead of GATK. Use bcftools for sample selection from multi-sample VCFs. Use pysam for calculation of sample coverage.
• Use GATK 3.0 MIT licensed framework for remaining BAM and variant manipulation code (PrintReads, CombineVariants) to provide one consistent up to date set of functionality for GATK variant manipulation.
• Normalize input variant_regions BED files to avoid overlapping segments. Avoids out of order errors with FreeBayes caller which will call in each region without flattening the input BED.

0.7.7 (February 27, 2014)

• For cancer tumor/normal calling, attach final call information of both to the tumor sample. This provides a single downstream file for processing and analysis.
• Enable batch specification in metadata to be a list, allowing a single normal BAM file to serve as a control for multiple tumor files.
• Re-organization of parallel framework code to enable alternative approaches. Document plugging in new parallel frameworks. Does not expose changes to users but makes the code cleaner for developers.
• Default to 1Gb/core memory usage when not specified in any programs. Do not use default baseline if supplied in input file. Thanks to James Porter.
• Integrate plotting of variant evaluation results using prettyplotlib.
• Add globals option to configuration to avoid needing to specify the same shared file multiple times in a samples configuration.
• Remove deprecated Celery distributed messaging, replaced in favor of IPython.
• Remove algorithm/custom_algorithm from bcbio_system.yaml, preferring to set these directly in the sample YAML files.
• Remove outdated and unused custom B-run trimming.
• Remove ability to guess fastq files from directories with no specification in sample YAML. Prefer using generalized template functionality with explicit specification of files in sample YAML file.
• Remove deprecated multiplex support, which is outdated and not maintained. Prefer approaches in external tools upstream of bcbio-nextgen.
• Add --tag argument which labels job names on a cluster to help distinguish when multiple bcbio jobs run concurrently. Thanks to Jason Corneveaux.
• Connect min_read_length parameter with read_through trimming in RNA-seq. Thanks to James Porter.
• Map variant calling specification to variant2 since original approach no longer supported.
• Fix issues with trying to upload directories to Galaxy. Thanks to Jim Peden.
• Made inner distance calculation for Tophat more accurate.
• Added gffutils GFF database to the RNA-seq indices.
• Add gene name annotation from the GFF file instead of from mygene.

0.7.6 (January 15, 2014)

• Expand template functionality to provide additional ability to add metadata to samples with input CSV. Includes customization of algorithm section and better matching of samples using input file names. Improve ability to distinguish fastq pairs.
• Generalize snpEff database preparation to use individual databases located with each genome. Enables better multi-organism support.
• Enable tumor/normal paired called with FreeBayes. Contributed by Luca Beltrame.
• Provide additional parallelization of bgzip preparation, performing grabix indexing in parallel for paired ends.
• Fix downsampling with GATK-lite 2.3.9 releases by moving to sambamba based downsampling. Thanks to Przemek Lyszkiewicz.
• Handle Illumina format input files for bwa-mem alignment, and cleanly convert these when preparing bgzipped inputs for parallel alignment. Thanks to Miika Ahdesmaki.
• Provide better algorithm for distinguishing bwa-mem and bwa-aln usage. Now does random sampling of first 2 million reads instead of taking the first set of reads which may be non-generalizable. Also lowers requirement to use bwa-mem to 75% of reads being smaller than 70bp. Thanks to Paul Tang.
• Enable specification of a GATK key file in the bcbio_system resources keyfile parameter. Disables callbacks to GATK tracking. Thanks to Severine Catreux for keyfile to debug with.
• Correctly handle preparation of pre-aligned BAM files when sorting and coordinate specification needed. Thanks to Severine Catreux.
• Fix incorrect quality flag being passed to Tophat. Thanks to Miika Ahdesmaki.
• Fix Tophat not respecting the existing --transcriptome-index. Thanks to Miika Ahdesmaki.
• Keep original gzipped fastq files. Thanks again to Miika Ahdesmaki.
• Fixed incompatibility with complexity calculation and IPython.
• Added strand-specific RNA-seq support via the strandedness option.
• Added Cufflinks support.
• Set stranded flag properly in htseq-count. Thanks to Miika Ahdesmaki.
• Fix to ensure Tophat receives a minimum of 8 gb of memory, regardless of number of cores.
• Remove hybrid_bait and hybrid_target which were no longer used with new lightweight QC framework. Prefer better coverage framework moving forward.
• Added extra summary information to the project-summary.yaml file so downstream tools can locate what genome resources were used.
• Added test_run option to the sample configuration file. Set it to True to run a small subset of your data through the pipeline to make sure everything is working okay.
• Fusion support added by setting fusion_mode: True in the algorithim section. Not officially documented for now until we can come up with best practices for it.
• STAR support re-enabled.
• Fixed issue with the complexity calculation throwing an exception when there were not enough reads.
• Add disambiguation stats to final project-summary.yaml file. Thanks to Miika Ahdesmaki.
• Remove Estimated Library Size and Complexity from RNA-seq QC summary information as they were confusing and unnecessarily alarming, respectively. Thanks to Miika Ahdesmaki and Sara Dempster.
• Several memory allocation errors resulting in jobs getting killed in cluster environments for overusing their memory limit fixed.
• Added JVM options by default to Picard to allocate enough memory for large BAM->FastQ conversion.

0.7.5 (November 29, 2013)

• Update overall project metrics summary to move to a flexible YAML format that handles multiple analysis types. Re-include target, duplication and variant metrics.
• Support disambiguation of mixed samples for RNA-seq pipelines. Handles alignment to two genomes, running disambiguation and continuation of disambiguated samples through the pipeline. Contributed by Miika Ahdesmaki and AstraZenenca.
• Handle specification of sex in metadata and correctly call X,Y and mitochondrial chromosomes.
• Fix issues with open file handles for large population runs. Ensure ZeroMQ contexts are closed and enable extension of ulimit soft file and user process limits within user available hard limits.
• Avoid calling in regions with excessively deep coverage. Reduces variant calling bottlenecks in repetitive regions with 25,000 or more reads.
• Improve bcbio_nextgen.py upgrade function to be more consistent on handling of code, tools and data. Now each require an implicit specification, while other options are remembered. Thanks to Jakub Nowacki.
• Generalize retrieval of RNA-seq resources (GTF files, transcriptome indexes) to use genome-resources.yaml. Updates all genome resources files. Contributed by James Porter.
• Use sambamba for indexing, which allows multicore indexing to speed up index creation on large BAM processing. Falls back to samtools index if not available.
• Remove custom Picard metrics runs and pdf generation. Eliminates dependencies on pdflatex and R for QC metrics.
• Improve memory handling by providing fallbacks during common memory intensive steps. Better handle memory on SLURM by explicitly allowing system memory in addition to that required for processing.
• Update fastqc runs to use a BAM files downsampled to 10 million reads to avoid excessive run times. Part of general speed up of QC step.
• Add Qualimap to generate plots and metrics for BAM alignments. Off by default due to speed issues.
• Improve handling of GATK version detection, including support for Appistry versions.
• Allow interruption of read_through trimming with Ctrl-C.
• Improve test suite: use system configuration instead of requiring test specific setup. Install and use a local version of nose using the installer provided Python.
• Fix for crash with single-end reads in read_through trimming.
• Added a library complexity calculation for RNA-seq libraries as a QC metric
• Added sorting via sambamba. Internally bcbio-nextgen now inspects the headers of SAM/BAM files to find their sorting status, so make sure tools set it correctly.

0.7.4 (October 20, 2013)

• Framework for indexing input reads using parallel bgzip and grabix, to handle distributed alignment. Enables further distribution of alignment step beyond multicore nodes.
• Rework of ensemble calling approach to generalize to population level ensemble calls. Provide improved defaults for handle 3 caller consolidation.
• Support for Mouse (mm10) variant calling and RNA-seq.
• For recent versions of Gemini (0.6.3+) do not load filtered variants into database, only including passed variants.
• Improve specification of resource parameters, using multiple -r flags instead of single semi-colon separated input. Allow specification of pename resource parameter for selecting correct SGE environment when not automatically found.
• Support biobambam's bammarkduplicates2 for duplicate removal.
• Clean up logging handling code to be more resilient to interrupt messages.
• Speed improvements for selecting unanalyzed and unmapped reads to address bottlenecks during BAM prep phase.
• Bug fix for algorithm options incorrectly expanded to paths on re-runs. Thanks to Brent Pedersen for report.
• Fix for Tophat 2.0.9 support: remove reads with empty read names.
• Save installation and upgrade details to enable cleaner upgrades without needing to respecify genomes, tool directory and other options from installation.

0.7.3 (September 22, 2013)

• Move specification of supporting genome files for variation (dbSNP, training files) and RNA-seq (transcript GTF files) analyses into an organism specific resources file. Improves ability to support additional organisms and genome builds.
• Provide paired tumor/normal variant calling with VarScan. Thanks to Luca Beltrame.
• Require bash shell and use of pipefail for piped commands. Ensures rapid detection of failures during piped steps like alignment.
• Use samtools cat for post-BAM merging to avoid issues with bamtools requirement for open file handles.
• Add installation/upgrade options to enable commercially restricted and data intensive third party tools.
• Support for GATK 2.7
• Fixes for TopHat 2.0.9 support: remove extra non-mate match paired end reads from alignment output.
• Pull description sample names from BAM files if not present in input configuration file. Thanks to Paul Tang for suggestion.
• Bug fixes for non-paired RNA-seq analysis.
• Add custom filtration of FreeBayes samples using bcbio.variation.
• Default to phred33 format for Tophat alignment if none specified.

0.7.2 (August 30, 2013)

• Report memory usage for processes to cluster schedulers and use predicted memory usage to schedule cores per machine. Gets core and memory information for machines and uses to ensure submitted jobs can schedule with available resources.
• Provide error checking of input YAML configuration at run start. Avoids accidental typos or incorrect settings that won't error out until later in the process.
• Drop requirement for fc_name and fc_date in input YAML file. Individual sample names are instead used and required to be unique within a processing run.
• Remove original variant pipeline, replacing with the all around better variant2 analysis method. Plan for the next version is to automatically redirect to variant2.
• Improve parallelization of BAM preparation and gemini database creation by moving to multicore versions.
• Move variant annotation to work on called sub-regions, to avoid bottlenecks when annotating a full whole genome VCF.
• Remove sequencer-specific integration functionality which is poorly maintained and better done with third party tools: demultiplexing and statistics from Illumina directories.
• Bug fix to re-enable template generation functionality.
• Improve BAM merging on large files using samtools for output sort.
• Uploading results works with the RNA-seq pipeline.
• Rework internals to provide a consistent dictionary of sample attributes up front, avoiding lane/sample dichotomy which provided confusing internal code.
• Drop calling htseq-count from the command line in favor of an internal implementation.

0.7.1 (August 12, 2013)

• Remove requirement for bcbio_system.yaml passed in on command line, defaulting to default file prepared by installer unless specified.
• Bug fixes for new approach to parsing *.loc files: handle Galaxy *.loc files with mixed tabs and spaces correctly and fall back to previous approaches when aligner specific *.loc files are missing.
• Bug fixes for preparing merged BAM files using bamtools: correctly sort after merging and avoid duplication of reads in noanalysis files.
• Bug fix for concatenating files when first file in empty.
• Recover from ZeroMQ logging errors, avoiding loss of logging output.

0.7.0 (July 30, 2013)

• RNA-seq pipeline updated: deprecate Tophat 1 in favor of Tophat 2. Perform automatic adapter trimming of common adapter sequences. STAR aligner support. RNA-SeQC support for RNA-seq specific quality control. Transcript quantitation with htseq-count.
• Updated installation and upgrade procedures, to make it easier to build an initial analysis pipeline and upgrade bcbio-nextgen and third-parts tools and data in place.
• Add support for MuTect tumor/normal variant caller, contributed by Luca Beltrame.
• Generalize variant calling to support alternative callers like cancer-specific calling: provide additional associated files to variant calls and pass along sample specific metadata. Document implementation of new variant callers.
• Improve algorithms around post-variant calling preparation. Avoid unnecessary tries for VQSR on low coverage whole genome reads, and concatenate VCF files to avoid locking penalties.
• Fix logging and memory usage for multicore jobs run within ipython clusters.
• Improve logging for IPython cluster issues, including moving IPython logs inside project logging directory for better access.
• Options for improved cluster resiliency: minimize number of clusters started during processing with more extensive reuse, flexible timeouts for waiting on cluster start up, and expose options to allow job retries. Thanks to Zhengqiu Cai for suggestions and testing.

0.6.5 (June 07, 2013)

• Improve logging: Detailed debugging logs collect all process standard out and error and command lines across distributed systems.
• Piping improvements: provide fully piped analysis with GATK recalibration and gkno realignment. Handle smaller reads with novoalign piped analysis.
• Improve collapsing analysis regions into evenly sized blocks to better handle large numbers of samples analyzed together.
• Provide template functionality to ease generation of input sample.yaml files from lists of BAM of fastq files. Thanks to Brent Pedersen and Paul Tang.
• Updated program support: Improved novoalign support based on evaluation with reference genomes. Support GATK 2.5-2. Support VarScan 2.3.5.
• Fix naming of read group information (ID and SM) to be more robust. Identifies issues with duplicated read groups up front to avoid downstream errors during variant calling. Thanks to Zhengqiu Cai.
• Improve quality control metrics: Cleanup into custom qc directory and ensure correct selection of duplicate and other metrics for split post-alignment prep, even without merging.
• Fix IPython parallel usage for larger clusters, providing improved resiliency for long running jobs.
• Clean up handling of missing programs and input files with better error messages. From Brent Pedersen.

0.6.4 (May 06, 2013)

• Integrate fully with bcbio.variation to provide automated validation of variant calls against reference materials.
• Provide full list of all third party software versions used in analysis.
• Create GEMINI database as part of output process, allowing immediate queries of variants with associated population and annotation data.
• Collapse analysis regions into evenly sized blocks separated by non-callable regions. Provides better parallelism.
• Documentation and examples for NA12878 exome and whole genome pipelines.