Browse files

Provide additional documentation on configuration parameters includin…

…g resources section and ensemble calling
  • Loading branch information...
chapmanb committed Jan 27, 2013
1 parent e9e0b9d commit 4d81bbadc3aa5050eae139bf4cbfcd518ad7916e
Showing with 111 additions and 9 deletions.
  1. +6 −1 config/bcbio_system.yaml
  2. +0 −1 config/examples/NA12878-ensemble.yaml
  3. +105 −7 docs/contents/configuration.rst
@@ -3,6 +3,8 @@
# These pipeline apply generally across multiple projects. Adjust them in sample
# specific configuration files when needed.
+# -- Base setup
# General attributes that apply across multiple pipelines.
aligner: bowtie
@@ -65,6 +67,10 @@ resources:
jvm_opts: ["-Xms2g", "-Xmx4g"]
dir: /usr/share/java/snpeff
+ bcbio_variation:
+ dir: /usr/share/java/bcbio_variation
+# -- Additional options for specific integration, not required for standalone usage.
# Galaxy integration. Required for upload/download from a Galaxy instance.
galaxy_url: http://your/galaxy/url
@@ -92,4 +98,3 @@ analysis:
base_dir: /array0/projects/Sequencing
@@ -22,7 +22,6 @@ details:
calling: [ReadPosEndDist, PL, PLratio, Entropy, NBQ]
type: svm
- classifier-type: svm
trusted-pct: 0.65
quality_format: Standard
coverage_interval: regional
@@ -17,12 +17,38 @@ Commented example files are available in the ``config`` directory:
- `example system config`_
- `example sample config`_
+Sample information
+The sample configuration file defines ``details`` of each sample to process::
+ details:
+ - analysis: variant
+ algorithm:
+ metadata:
+ batch: Batch1
+ description: Example1
+ genome_build: hg19
+- ``analysis`` Analysis method to use [variant, RNA-seq]
+- ``algorithm`` Parameters to configure algorithm inputs. Options
+ described in more detail below.
+- ``metadata`` Additional descriptive metadata about the sample. The
+ ``batch`` input defines a batch that the sample falls in. We perform
+ multi-sample variant calling on all samples with the same batch name.
+- ``description`` Unique name for this sample. Required.
+- ``genome_build`` Genome build to align to, which references a genome
+ keyword in Galaxy to find location build files.
+Algorithm parameters
The YAML configuration file provides a number of hooks to customize
analysis in the sample configuration file. Place these under the
-``analysis`` keyword. For variant calling:
+``analysis`` keyword.
- ``aligner`` Aligner to use: [bwa, bowtie, bowtie2, mosaik, novoalign,
@@ -31,21 +57,31 @@ analysis in the sample configuration file. Place these under the
- ``align_split_size``: Split FASTQ files into specified number of
records per file. Allows parallelization at the cost of increased
temporary disk space usage.
-- ``variantcaller`` Variant calling algorithm [gatk, freebayes]
- ``quality_format`` Quality format of fastq inputs [illumina,
+- ``write_summary`` Write a PDF summary of results [true, false]
+Experimental information
- ``coverage_interval`` Regions covered by sequencing. Influences GATK
options for filtering [exome, genome, regional]
- ``coverage_depth`` Depth of sequencing coverage. Influences GATK
variant calling [high, low]
- ``hybrid_target`` BED file with target regions for hybrid selection
-- ``variant_regions`` BED file of regions to call variants in.
- ``ploidy`` Ploidy of called reads. Defaults to 2 (diploid).
-- ``recalibrate`` Perform variant recalibration [true, false]
+Variant calling
+- ``variantcaller`` Variant calling algorithm. Can be a list of
+ multiple options [gatk, freebayes, varscan, samtools,
+ gatk-haplotype, cortex]
+- ``variant_regions`` BED file of regions to call variants in.
- ``mark_duplicates`` Identify and remove variants [false, true]
+- ``recalibrate`` Perform variant recalibration [true, false]
- ``realign`` Do variant realignment [true, false]
-- ``write_summary`` Write a PDF summary of results [true, false]
Broad's `GATK`_ pipeline drives variant (SNP and Indel) analysis.
This requires some associated data files, and also has some configurable
@@ -61,6 +97,68 @@ are inputs into the training models for recalibration. The automated
`CloudBioLinux`_ data scripts will download and install these in the
variation subdirectory relative to the genome files.
+Ensemble variant calling
+In addition to single method variant calling, we support calling with
+multiple calling methods and consolidating into a final Ensemble
+callset. This requires the `bcbio.variation`_ toolkit to perform the
+consolidation. An example configuration in the ``algorithm`` section is::
+ variantcaller: [gatk, freebayes, samtools, gatk-haplotype, varscan]
+ ensemble:
+ format-filters: [DP < 4]
+ classifier-params:
+ type: svm
+ classifiers:
+ balance: [AD, FS, Entropy]
+ calling: [ReadPosEndDist, PL, PLratio, Entropy, NBQ]
+ trusted-pct: 0.65
+The ``ensemble`` set of parameters configure how to combine calls from
+the multiple methods:
+- ``format-filters`` A set of filters to apply to variants before
+ combining. The example removes all calls with a depth of less than
+ 4.
+- ``classifier-params`` Parameters to configure the machine learning
+ approaches used to consolidate calls. The example defines an SVM
+ classifier.
+- ``classifiers`` Groups of classifiers to use for training and
+ evaluating during machine learning. The example defines two set of
+ criteria for distinguishing reads with allele balance issues and
+ those with low calling support.
+- ``trusted-pct`` Define threshold of variants to include in final
+ callset. In the example, variants called by more than 65% of the
+ approaches (4 or more callers) pass without being requiring SVM
+ filtering.
+The ``resources`` section allows customization of locations of programs
+and memory and compute resources to devote to them::
+ resources:
+ bwa:
+ cores: 12
+ cmd: /an/alternative/path/to/bwa
+ gatk:
+ jvm_opts: ["-Xms2g", "-Xmx4g"]
+ dir: /usr/share/java/gatk
+- ``cmd`` Location of an executable. By default, we assume executables
+ are on the path.
+- ``dir`` For software not distributed as a single executable, like
+ files of Java jars, the location of the base directory.
+- ``cores`` Cores to use for multi-proccessor enabled software.
+- ``jvm_opts`` Specific memory usage options for Java software.
+Resources will continue to expand to allow direct customization of
+commandline options as well as fine grained control over research
+.. _bcbio.variation:
.. _CloudBioLinux:
.. _YAML format:
.. _GATK resource bundle:

0 comments on commit 4d81bba

Please sign in to comment.