chapmanb/bcbio-nextgen

Thanks to @lpantano for pointing out the issue.

This lets us know exactly what we quantitated against for a
particular project.

Reworks java path detection and usage to try and ensure we use
anaconda installed Java 8. Fixes #2029

Do conversion to CRAM without quality score binning. Thanks to
Alison Meynert.

Closes #2025. Thanks to Rod Docker (@rdocking) for the great bug
report and finding out the solution.

Work around issues on machines with multiple IPs and no valid way
to resolve by initially checking, then falling back to previous
approaches if it's a local reference.

Fixes #2027

bioconda plink is 1.90 but the exectuable is named plink2.

Replaces cutadapt as default read trimmer. atropos
(https://atropos.readthedocs.io/en/latest/) is a drop in
replacement with multithreading and streaming outputs to bgzip. bcbio
implementation uses the same paramters as previous cutadapt approach.

- Removes extra trim and double grabix prep steps for alignment
  pre-trimming, avoiding IPython issues.
- Removes old align_prep_method configuration parameter.

Replaces sambamba indexing with new native samtools multicore indexing,
fixed in 1.5 release (samtools/samtools#687). Avoids intermittent
system specific segfaults with sambamba.

Add multicore reading introduced in beta3 release.

Swap to using vcfanno instead of GEMINI for retrieving metrics for
tumor-only prioritization and germline estimation. Avoids needing
a full GEMINI database for assessment.

Parallel BQSR in GATK4 requires 2bit formatted genomes. Provide slot
and correct lookup of these in preparation for installing them
as part of bcbio.

Avoids a resource calculation error before the code that correctly
identifies and provides a useful error message for this error.
Fixes #2024

- Adds GATK4 GatherVcfs, a new rapid way in GATK to combine VCFs
  similar to bcftools concat --naive. Fixes #2023
- fixes issues with initial scattered contigs from chrM. These are
  haploid called so don't have PID phasing key/value pairs, causing
  missing header problems. Also fixes #1999 which has key errors
  due to checking for GATK variantcallers, which we no longer need
  to do.

Stream to maximum coverage downsampling during merging for align_split
samples or after sorting for non split. Avoids slow runtimes for
separate BAM downsampling. Fixes #2022

Closes #1709. Thanks to @schelhorn for the suggestion.

Thanks to @NeillGibson. Fixes #2021

- Document usage of GATK4 with tools on
- Add support for VQSR, updating to latest command line
- Avoid problematic MuTect2 BaseQualityRankSumTest annotation with
  latest GATK4 beta.2 release.

Trying to get TravisCI working again with less space for tests
(no STAR index for mm9).

For super deep libraries, the default wasn't inadequate. Thanks
to Manasa Surakala for pointing out the problem.

Each STAR index uses ~1.5 GB, which was mucking up the ability to
do automatic tests.