SOP_alignment_variant_annotation.md

Standard operating procedure - Alignment, variant calling, and annotation of trio whole exome samples at the Edinburgh Parallel Computing Centre

This SOP applies to batches of family/trio samples where trio whole exome sequencing has been performed by Edinburgh Genomics (EdGE) or the Edinburgh Clinical Research Facility (ECRF). It assumes that data has been successfully transferred to the Edinburgh Parallel Computing Centre (EPCC) (see SOP: Transfer of whole exome sequencing samples from Edinburgh Genomics to Edinburgh Parallel Computing Centre). Scripts are version controlled on the University of Edinburgh gitlab server gitlab.ecdf.ed.ac.uk/igmmbioinformatics/trio-whole-exome. Request access by e-mail: alison.meynert@igmm.ed.ac.uk.

Definitions

In this document, N is the total number of samples in the project, and X is the number of families.

Text in angle brackets, e.g. <project> indicates variable parameters. A variable parameter such as <family1-X> indicates that there are X instances of the parameter, each with their own unique value.

Software and data requirements

The analysis is run with the bcbio pipeline (version 1.2.8) located at /home/u035/u035/shared/software/bcbio. All genome reference and annotation data resources are contained within the genomes/Hsapiens/hg38 subfolder.

The TWIST target BED file is at: /home/u035/u035/shared/resources/Twist_Exome_RefSeq_targets_hg38.plus15bp.bed. See resources.

Input

PED file

A 6-column tab-delimited PED/FAM format file is required for each batch, describing the relationships between the sampled individuals, their sex, and their affected/unaffected status.

Sample id format

The sequencing reads for the samples delivered from EdGE are identified by folder name and as the 8th column in the tab-delimited text file file_list.tsv inside the dated batch directory. The identifiers are in the format:

<pcr_plate_id>_<indiv_id>_<family_id><suffix>

The suffix identifies the exome kit, e.g. _WESTwist_IDT-A. These identifiers are referenced below in the output file structure.

Reads - Edinburgh Genomics

A set of paired end FASTQ files (designated by R1 or R2 suffixes), possibly more than one pair per sample. Each sample's files are in its own folder. The input files will be in the folder /home/u035/u035/shared/data and follow the structure in Figure 1. Older deliveries contained the <dated_batch> folder within a raw_data folder.

<EdGE_project_id>/
  +---<dated_batch>/
  |   +---<sample_id>/
  |   |   +---*.fastq.count
  |   |   +---*.fastq.gz
  |   +---file_list.tsv
  |   +---md5sums.txt
  +---<dated_batch>_tree.txt
  +---Information.txt
  +---md5_check.txt

Figure 1. File name and directory structure for a batch of sequencing from Edinburgh Genomics. The EdGE project id takes the format XXXXX_Lastname_Firstname, identifying the NHS staff member who submitted the samples for sequencing. The dated batch is in the format yyyymmdd – in general we expect there to be only one of these per EdGE project id. The FASTQ file names relate to the sequencing run information and do not contain any information about the sample itself.

Reads - Edinburgh Clinical Research Facility

A set of paired end FASTQ files (designated by R1 or R2 suffixes), generally one pair per sample. The input files will be in the folder /home/u035/u035/shared/data and follow the structure in Figure 2.

<ECRF_project_id>/
  +---<internal_id_-md5.txt
  +---<pcr_plate_id>_<indiv_id>_<family_id><suffix>_S<i>_L001_R1_001.fastq.gz
  +---<pcr_plate_id>_<indiv_id>_<family_id><suffix>_S<i>_L001_R2_001.fastq.gz
  +...

Figure 2. File name and directory structure for a batch of sequencing from the ECRF.

Working directories

The project working directories will be in the folder /home/u035/u035/shared/analysis and follow the structure in Figure 3.

    config – bcbio configuration files in YAML format
    logs – PBS job submission log files
    params – parameters for PBS job submission
    reads – symlinks/merged versions of input FASTQ files
    work – bcbio working folder

Figure 3. Project working directories.

Project configuration

A configuration script sets environment variables common to scripts used in this SOP.

Template for bcbio configuration

Bcbio requires a template file in YAML format to define the procedures run in the pipeline.

Output

Per sample: BAM file of aligned reads against the hg38 genome assembly Per family: Annotated VCF file and QC report

Output will be in the folder /home/u035/u035/shared/results/<short_project_id>_<version> where <short_project_id> is the numeric prefix of <project_id> and follow the structure in Figure 4 (with multiple instances of the indiv_id sub directories, one per sequenced family member.). The qc sub-directories are not enumerated, and automatically generated index files are not listed for brevity. An additional directory at the root of each project/version output folder called “qc” will contain the MultiQC report generated for an entire batch.

<analysis_date>_<project_id>_<pcr_plate_id>_<family_id>/
  +---<indiv_id>_<family_id>/
  |   +---<indiv_id>_<family_id>-callable.bed
  |   +---<indiv_id>_<family_id>-ready.bam
  |   +---qc/
  +---<pcr_plate>_<family_id>-gatk-haplotype-annotated.vcf.gz
  +---bcbio-nextgen-commands.log
  +---bcbio-nextgen.log
  +---data_versions.csv
  +---metadata.csv
  +---multiqc/
  |   +---list_files_final.txt
  |   +---multiqc_config.yaml
  |   +---multiqc_data/
  |   +---multiqc_report.html
  |   +---report/
  +---programs.txt
  +---project-summary.yaml

Figure 4. File name and output directory structure for each family in a batch of sequencing.

Procedure

1. Set environment variable project_id and general configuration variables. All steps below can assume that these have been set. Version should be "v1" by default for
   the first analysis run of a batch, "v2" etc for subsequent runs.

project_id=<project_id>
short_project_id=`echo $project_id | cut -f 1 -d '_'`
version=<version>.
source /home/u035/u035/shared/scripts/trio_whole_exome_config.sh

2. Copy the PED file for the batch to the params folder in the working area. It should be named <project_id>.ped, relating it to the input directory for the FASTQ files. If the PED file given was not named in this way, don’t rename it, copy it instead.

cd $PARAMS_DIR
ped_file=<input_ped_file>
cp $ped_file $project_id.ped

3. In the params folder, create the symlinks to the reads and the bcbio configuration files. If specifying a common sample suffix, ensure it includes any joining characters, e.g. “-“ or “_”, so that the family identifier can be cleanly separated from the suffix. Get the number of families from the batch.

Edinburgh Genomics data

cd $PARAMS_DIR
sample_suffix=<sample_suffix>
$SCRIPTS/prepare_bcbio_config.sh \
  $SCRIPTS/trio_whole_exome_config.sh \
  $project_id $version $sample_suffix &> ${project_id}_${version}_`date +%Y%m%d%H%M`.log
X=`wc -l $PARAMS_DIR/$project_id.family_ids.txt | awk '{print $1}'`

Edinburgh Clinical Research Facility data

cd $PARAMS_DIR
sample_suffix=<sample_suffix>
$SCRIPTS/scripts/prepare_bcbio_config_crf.sh \
  $SCRIPTS/trio_whole_exome_crf_config.sh \
  $project_id $version $sample_suffix &> ${project_id}_${version}_`date +%Y%m%d%H%M`.log
X=`wc -l $PARAMS_DIR/$project_id.family_ids.txt | awk '{print $1}'`

4. Submit the bcbio jobs from the logs folder. See above for version.

cd $LOGS_DIR
sbatch --export=PROJECT_ID=$project_id,VERSION=$version,CONFIG_SH=$SCRIPTS/trio_whole_exome_config.sh \
  --array=1-$X --job-name=trio_whole_exome_bcbio.$short_project_id \
  $SCRIPTS/submit_bcbio_trio_wes.sh

If all log files end in ‘Finished’ or ‘Storing in local filesystem’ for a metadata file (occasionally the job completes without quite outputting all of the ‘Storing’ messages), the batch is complete. If this is not the case, resubmit the incomplete jobs – they will resume where they left off.

5. Check the output directory to make sure all family output folders were moved into the families subdirectory. This should happen automatically at the end of the submit_bcbio_trio_wes.sh script but occasionally fails.

cd $OUTPUT_DIR/${short_project_id}_${version}
mv *${short_project_id}* families/

6. Generate a MultiQC report for all files in the batch.

cd $OUTPUT_DIR/${short_project_id}_${version}/families
mkdir -p ../qc
multiqc --title "Trio whole exome QC report: $short_project_id $version" \
  --outdir ../qc \
  --filename ${short_project_id}_${version}_qc_report.html .

7. Check the parent-child relationships predicted by peddy match the pedigree information. There should be no entries in the <EdGE_project_id>.ped_check.txt file that do not end in ‘True’. If there are, report these back to the NHS Clinical Scientist who generated the PED file for this batch. The <batch_id> is the 5 digit number that prefixes all the family ids in the output.

cd $OUTPUT_DIR/${short_project_id}_${version}/families
batch_id=<batch_id>

perl $SCRIPTS/trio_whole_exome_parse_peddy_ped_csv.pl \
  --output ../qc/${short_project_id}_${version}.ped_check.txt \
  --project $project_id \
  --batch $batch_id \
  --version $version \
  --ped $PARAMS_DIR/$project_id.ped
grep -v False$ ../qc/${short_project_id}_${version}.ped_check.txt

8. Clear the work and logs directories.

cd $WORK_DIR
rm -r *
cd $LOGS_DIR
rm -r *

9. Clean up the reads directory. Retain reads for samples in families where one sample has failed QC, using a list retain\_for\_rerun.txt. These will likely be required for later runs, and it is simpler to regenerate config YAML files if it is not necessary to re-do symlinks/read merging.

cd /home/u035/u035/shared/analysis/reads/${project_id}
rm `ls | grep -v -f retain_for_rerun.txt`

10. Update the batch status spreadsheet. Move to SOP prioritization.