UGP Variant Pipeline Sandbox

From Utah Genome Project Wiki
Jump to navigation Jump to search

Sources of Analysis Info

DNA Samples

The starting point for sample prep guidelines is the HCI Bioinfo/Genomics Cores Wiki sample prep page. Refer to that page and we can lock down details here as that becomes appropriate.


Illumina HiSeq paired-end 101 bp sequencing (with an insert size of 200-500??)

Data that we will need from sequencing and library prep for readgroup designation later on. See the SAM format page for details:

  • ID: Read group identifier.
  • CN: Name of sequencing center producing the read.
  • DS: Description.
  • DT: Date the run was produced (ISO8601 date or date/time).
  • FO: Flow order. The array of nucleotide bases that correspond to the nucleotides used for each flow of each read.
  • KS: The array of nucleotide bases that correspond to the key sequence of each read.
  • LB: Library.
  • PG: Programs used for processing the read group.
  • PI: Predicted median insert size.
  • PL: Platform/technology used to produce the reads. One of (CAPILLARY, LS454, ILLUMINA, SOLID, HELICOS, IONTORRENT, PACBIO).
  • PU: Platform unit (e.g. owcell-barcode.lane for Illumina or slide for SOLiD). Unique identifier.
  • SM: Sample. Use pool name where a pool is being sequenced.

FastQ File Analyses

Validate File Integrity with md5sum

An md5sum signature should be provided for each FastQ file by the group providing the sequencing services. After the file has been downloaded locally check the md5sum to be sure no data corruption occurred during the file transfer.

md5sum file.fastq > file.md5
diff file.md5 provided.md5

If the md5sum signature differs from that provided for the file:

  • Check to be sure you have the correct file.
  • Check if the md5sum was calculated on that compressed or uncompressed file by the provider and be sure to do the same with the local copy.
  • Try the download again.
  • Contact the sequence provider.

Check Sequence Quality with FastQC

FastQC is a Java utility that calculates a number of metrics that are useful in assessing the quality of sequence data.

fastqc 9685X1_130201_SN1117_0141_BC1JBVACXX_1_1.txt

FastQC provides the following metrics:

  • Basic Statistics
  • Per Base Sequence Quality
  • Per Sequence Quality Score
  • Per Base Sequence Content
  • Per Base GC Content
  • Per Base N Content
  • Sequence Length Distribution
  • Sequence Duplication Levels
  • Overrepresented Sequences
  • Kmer Content

Convert Quality Scores to Sanger Standard

Base calling qualities are represented in FastQ files as an ASCII representation of Phred-scaled quality scores. Different applications have used different transformations to generate ASCII codes for quality scores. Most if not all aligners interpret ASCII representations of the quality score based on the Sanger Standard. The Basic Statistics section of the FastQC output will provide an 'Encoding' value for your file. If this value is anything other than 'Sanger / Illumina 1.9', then you will either need to map the quality values in your FastQ files to the Sanger standard, or use a command line option to your aligner to let it know the correct encoding.

Other Possible Analyses

Several other tools exist that can provide custom analyses of FastQ files - you may want to consider the following:

  • fastq_tool: This is a perl script written and maintained by Barry. Most of the functionality of this tool has now been included in faster packages like FastQC above. However, the script is fast enough to burn through a large Fastq file in a couple of hours, so if you need to write a custom analysis quickly and don't have a lot of files to process this may still be the path of least resistance. The script is included in the GAL library.
  • FASTX is an application written by the Hannon Lab at CSHL. It provides a number of tools for calculating quality metrics and manipulating Fastq files.


Which Aligner to use?

  • bwa: Written and maintained by Heng Li while with Richard Durbin's group at the Sanger Institute. It has become the defacto standard aligner for NGS-based variant discovery pipelines due in large part to it's leading role in analyses by the 1000 Genomes project. It will be the aligner described below, however the merits of the other aligners should be carefully considered.
  • Bowtie2: Bowtie2 is developed and maintained by Steve Salsberg's group at Johns Hopkins University. It is the major competitor to bwa and has developed a strong following.
  • Mosaic: Written and maintained by Gabor Marth's group at Boston College, Mosaic produces gapped alignments using the Smith-Waterman algorithm which means that it should do a better job at indel discovery and prehaps do better at avoiding false-positive SNV calls in the vicinity of indels.

What reference assembly to use?

The first consideration when aligning NGS data with any aligner is, "What reference sequence will the reads be aligned to?". The answer is of course, "Use the current human reference assembly", which at the time of this writing was GRCh37.p12 (also referred to as hg19 by the UCSC Genome Browser). However, it is important to include decoy sequences with the reference assembly because they can prevent reads that represent sequence not found in the reference assembly (retro-viral sequence, unanchored contigs, alternate haplotypes etc) from mis-aligning to the reference and creating false positive variant calls. The 1000 Genomes project has put together such a baited reference assembly Fasta file. This is the primary file to use for alignments - particularly if the 1000 Genomes data will be used for comparison later on in the pipeline. The file, as well as a document describing it's construction can be found with the 1KG technical reference data at:

Alignment of Illumina HiSeq PE 101 bp sequences with bwa aln

The first step is to index the reference sequence.

bwa index -a bwtsw $reference_fasta

The first alignment step will find the reference coordinates of the input reads (independent of their mate-pair). The bwa aligner has many options which can control various aspects of the alignment. For the most-part we use default parameters for the alignment and this will likely suffice for most circumstances, however the command line parameters should be carefully considered at the start of each project. See the bwa manual page and the [[Bwa | Yandbeck Lab Wiki page on bwa] for guidance in choosing appropriate command-line parameters. The following parameters are those used by the 1KG project for aligning Illumina data.

bwa aln -q 15 reference.fasta file.fastq > output.sai

Next create SAM files using bwa sampe. For paired-end reads, the maximum insert size is taken to be 3 times the expected insert size.

bwa sampe -P reference.fasta output1.sai output2.sai file1.fastq file2.fastq > out.sam

Command Line Options:

  • -n NUM max #diff (int) or missing prob under 0.02 err rate (float) [0.04]
  • -o INT maximum number or fraction of gap opens [1]
  • -e INT maximum number of gap extensions, -1 for disabling long gaps [-1]
  • -i INT do not put an indel within INT bp towards the ends [5]
  • -d INT maximum occurrences for extending a long deletion [10]
  • -l INT seed length [32]
  • -k INT maximum differences in the seed [2]
  • -m INT maximum entries in the queue [2000000]
  • -t INT number of threads [1]
  • -M INT mismatch penalty [3]
  • -O INT gap open penalty [11]
  • -E INT gap extension penalty [4]
  • -R INT stop searching when there are >INT equally best hits [30]
  • -q INT quality threshold for read trimming down to 35bp [0]
  • -c input sequences are in the color space
  • -L log-scaled gap penalty for long deletions
  • -N non-iterative mode: search for all n-difference hits (slooow)
  • -f FILE file to write output to instead of stdout

Alignment of Illumina HiSeq PE 101 bp sequences with bwa mem

      -t INT     number of threads [1]
      -k INT     minimum seed length [19]
      -w INT     band width for banded alignment [100]
      -d INT     off-diagonal X-dropoff [100]
      -r FLOAT   look for internal seeds inside a seed longer than {-k} * FLOAT [1.5]
      -c INT     skip seeds with more than INT occurrences [10000]
      -P         skip pairing; perform mate SW only
      -A INT     score for a sequence match [1]
      -B INT     penalty for a mismatch [4]
      -O INT     gap open penalty [6]
      -E INT     gap extension penalty; a gap of size k cost {-O} + {-E}*k [1]
      -L INT     penalty for clipping [5]
      -U INT     penalty for an unpaired read pair [9]

Input/output options:

      -p         first query file consists of interleaved paired-end sequences
      -R STR     read group header line such as '@RG\tID:foo\tSM:bar' [null]
      -v INT     verbose level: 1=error, 2=warning, 3=message, 4+=debugging [3]
      -T INT     minimum score to output [30]
      -a         output all alignments for SE or unpaired PE
      -C         append FASTA/FASTQ comment to SAM output
      -H         hard clipping
      -M         mark shorter split hits as secondary (for Picard/GATK compatibility)

BAM File Analyses

Alignment BAM files are improved in various ways to help increase the quality and speed of subsequent variant calling.

The 1KG pipeline has the following:

samtools view -bSu $sam_file | samtools sort -n -o - samtools_nsort_tmp |         \
samtools fixmate /dev/stdin /dev/stdout | samtools sort -o - samtools_csort_tmp | \
samtools fillmd -u - reference.fasta > fixed_sorted.bam

However GATK takes care of both the BAQ calculation and the fixmate step, so the only thing needed now is adding read groups and sorting. It's also best to switch to Broad tools (Picard/GATK) now if you plan to use them downstream, because subtle incompatibilities of BAM bewteen samtools and Picard/GATK can bite you downstream.

nohup java -d64 -Xmx8g -jar /usr/local/picard-tools-1.88/AddOrReplaceReadGroups.jar \
   INPUT=9958X1_7_1.bam                                                            \
   OUTPUT=9958X1_RG.bam                                                            \
   CREATE_INDEX=TRUE                                                               \
   MAX_RECORDS_IN_RAM=2000000                                                      \
   VALIDATION_STRINGENCY=SILENT                                                    \
   TMP_DIR=/tmp                                                                    \
   SORT_ORDER=coordinate                                                           \
   RGID=9958X1                                                                     \
   RGLB=9958X1                                                                     \
   RGPL=illumina                                                                   \
   RGPU=1                                                                          \
   2> 9958X1_AddOrReplaceReadGroups.error &

TODO: Should probably merge all lanes per individual here so that MarkDuplicates and IndelRealigner does their best work.

Merge lane level BAMs to individual

java -jar -Xmx4g /tools/picard-tools-1.90/MergeSamFiles.jar \                                                                                              
    INPUT=alignment.bam          \                                                                                                                         
    INPUT=alignment.bam          \                                                                                                                         
    OUTPUT=merged.bam            \                                                                                                                         
    CREATE_INDEX=TRUE            \                                                                                                                         
    ASSUME_SORTED=true           \                                                                                                                         
    USE_THREADING=true           \                                                                                                                         
    TMP_DIR=/tmp                 \                                                                                                                         
    MAX_RECORDS_IN_RAM=30000000  \                                                                                                                         
    SORT_ORDER=coordinate        \                                                                                                                         

GATK Local Realignment of Indels

java -Xmx10g -jar /usr/local/GenomeAnalysisTK-2.4-7-g5e89f01//GenomeAnalysisTK.jar                         \
   -I all_bams.list                             							   \
   -T RealignerTargetCreator                    							   \
   -R fasta/hs_ref_GRCh37.p10_ALL_chr_hs37d5.fa 							   \
   -o realign.intervals                    							           \
   -known known_indels/ALL.wgs.indels_mills_devine_hg19_leftAligned_collapsed_double_hit.indels.sites.vcf \
   -known known_indels/ALL.wgs.low_coverage_vqsr.20101123.indels.sites.vcf                                \
   -nt 24                                                                                                 \
   >  RealignerTargetCreator                                                                              \
   2> RealignerTargetCreator.error &
java -Xmx10g -jar /usr/local/GenomeAnalysisTK-2.4-7-g5e89f01/GenomeAnalysisTK.jar                          \
   -T IndelRealigner                                                                                      \
   -R fasta/hs_ref_GRCh37.p10_ALL_chr_hs37d5.fa                                                           \
   -I 9958X1_RG.bam                                                                                       \
   -I 9958X2_RG.bam                                                                                       \
   -I 9958X3_RG.bam                                                                                       \
   -I 9958X4_RG.bam                                                                                       \
   -targetIntervals realign.intervals                                                                     \
   -o 9958X_Realigned.bam                                                                                 \
   -known known_indels/ALL.wgs.indels_mills_devine_hg19_leftAligned_collapsed_double_hit.indels.sites.vcf \
   -known known_indels/ALL.wgs.low_coverage_vqsr.20101123.indels.sites.vcf                                \
   >  IndelRealigner.out                                                                                  \
   2> IndelRealigner.error

Or this ??:

java -Xmx8g -jar GenomeAnalysisTK.jar \
   -l INFO -T IndelRealigner \
   -I /output/filename.b37_1kg.dedup.bam \
   -targetIntervals /resources//hg19/intervals/realign_intervals_hg19_b37_1kg.intervals \
   -R /resources//hg19/indices/b37_1kg.fa \
   -D /resources//hg19/dbsnp/dbsnp_129_b37_b37_1kg.rod \
   -[B:indels,VCF B:indels,VCF] /resources//hg19/indels/1kg.pilot_release.merged.indels.sites./hg19.b37_1kg.vcf \
   -o /output/filename.b37_1kg.realigned.bam
   -LOD 0.4
   -compress 0

Known indels are from the following files:

Fix Mate Information

java -jar -Xmx6g /tools/picard-tools-1.32/FixMateInformation.jar \
   INPUT=/output/filename.b37_1kg.realigned.bam \
   OUTPUT=/output/filename.b37_1kg.matefixed.bam \
   SORT_ORDER=coordinate                       \

GATK Quality Score Recalibration

java [jvm_args] -jar GenomeAnalysisTK.jar                                 \
     -T CountCovariates                                                   \
     -R $reference_fasta                                                  \
     -I $realigned_bam_file                                                \
     -recalFile recal_data.csv                                            \
     -knownSites $known_sites_file(s)                                      \
     -l INFO                                                              \
     -L '1;2;3;4;5;6;7;8;9;10;11;12;13;14;15;16;17;18;19;20;21;22;X;Y;MT' \
     -cov ReadGroupCovariate                                              \
     -cov QualityScoreCovariate                                           \
     -cov CycleCovariate                                                  \
     -cov DinucCovariate                                                  \
java [jvm_args] -jar GenomeAnalysisTK.jar \ 
     -T TableRecalibration                \
     -R $reference_fasta                  \
     -recalFile recal_data.csv            \
     -I $realigned_bam_file                \
     -o $recalibrated_bam_file             \
     -l INFO                              \
     -compress 0                          \
     -disable_bam_indexing                \

Known sites for recalibration are from dbSNP135:

Sort and index recalibrated alignment (~5h)

java -jar -Xmx3g /tools/picard-tools-1.32/SortSam.jar \
   INPUT=/output/filename.b37_1kg.recal.bam            \
   OUTPUT=/output/filename.b37_1kg.recal.sorted.bam    \
   SORT_ORDER=coordinate                              \
   VALIDATION_STRINGENCY=LENIENT                      \
java -jar -Xmx3g /tools/picard-tools-1.32/BuildBamIndex.jar \
   INPUT=/output/filename.b37_1kg.recal.sorted.bam           \
   OUTPUT=/output/filename.b37_1kg.recal.sorted.bam.bai      \

Calculate covariates after realignment and recalibration

java -jar -Xmx2g GenomeAnalysisTK.jar                      \
   -l INFO                                                 \
   -T CountCovariates                                      \
   -U ALLOW_UNINDEXED_BAM                                  \
   -R /resources//hg19/indices/b37_1kg.fa                  \
   --DBSNP /resources//hg19/dbsnp/dbsnp_129_b37_b37_1kg.rod \
   -I /output/filename.b37_1kg.recal.sorted.bam             \
   -cov ReadGroupcovariate                                 \
   -cov QualityScoreCovariate                              \
   -cov CycleCovariate                                     \
   -cov DinucCovariate                                     \
   -recalFile /output/filename.b37_1kg.recal.covariate_table.csv

Analyze covariates before and after

java -jar -Xmx4g AnalyzeCovariates.jar                             \
   -l INFO                                                         \
   -resources /resources//hg19/indices/b37_1kg.fa                  \
   --recal_file /output/filename.b37_1kg.matefixed.covariate_table.csv \
   -outputDir /output/filename.b37_1kg.recal.stats_before/          \
   -Rscript ${rscript}
   -ignoreQ 5
java -jar -Xmx4g AnalyzeCovariates.jar                          \
   -l INFO                                                      \
   -resources /resources//hg19/indices/b37_1kg.fa               \
   --recal_file /output/filename.b37_1kg.recal.covariate_table.csv \
   -outputDir /output/filename.b37_1kg.recal.stats_after/        \
   -Rscript ${rscript}                                          \
   -ignoreQ 5

Generate the SAM MD Tag

The MD tag in SAM files provides a string representation of mismatching positions. The CIGAR string used in SAM files does not provide full detail of mismatch positions. The samtools calmd command is used to generate MD tags as well as which fix the NM tags (Edit distance to the reference) and introduce the BQ tags (Base Alignment Quality) which can be used during variant calling.

samtools calmd -Erb recalibrated_file.bam $reference.fasta > $bq_file.bam

Strip Extraneous BAM Tags

Run-level BAMs have extraneous tags (OQ, XM, XG, XO) stripped from them, to reduce total file size by around 30%.

GATK ReduceReads ???

Merge BAMS

Merge all BAMS for a given individual

java $jvm_args -jar MergeSamFiles.jar INPUT=$tag_stripped_bam_file(s) OUTPUT=$library_level_bam VALIDATION_STRINGENCY=SILENT

Mark PCR Duplicates

Picard MarkDuplicates is used to mark PCR Duplicates

java $jvm_args -jar MarkDuplicates.jar INPUT=$library_level_bam OUTPUT=$markdup_bam_file ASSUME_SORTED=TRUE METRICS_FILE=/dev/null VALIDATION_STRINGENCY=SILENT
java -Xmx4g -jar /tools/picard-tools-1.32/MarkDuplicates.jar \ INPUT=/output/filename.b37_1kg.sorted.bam \ OUTPUT=/output/filename.b37_1kg.dedup.bam \ METRICS_FILE=/output/filename.b37_1kg.dedup.metrics \ REMOVE_DUPLICATES=false ASSUME_SORTED=true VALIDATION_STRINGENCY=LENIENT \ TMP_DIR=/local

Merge All BAMS for the Project

This happens above... Picard MergeSamFiles is used to Merge all the SAM files for a given project.

java $jvm_args -jar MergeSamFiles.jar INPUT=$markdup_bam_file(s) OUTPUT=$platform_level_bam VALIDATION_STRINGENCY=SILENT

BAM Quality Control

  • Picard Tools
    • BamIndexStats
    • CalculateHsMetrics ??
    • CleanSam
    • CollectAlignmentSummaryMetrics
    • CollectGcBiasMetrics
    • CollectInsertSizeMetrics
    • CollectMultipleMetrics
    • CollectTargetedPcrMetrics
    • EstimateLibraryComplexity
    • FixMateInformation
    • MarkDuplicates
    • MeanQualityByCycle
    • QualityScoreDistribution
    • ValidateSamFile
  • [1]
  • [2]


java -jar -Xmx4g /tools/picard-tools-1.32/CollectAlignmentSummaryMetrics.jar \
   I=/output/filename.b37_1kg.sorted.bam                                      \
   O=/output/filename.b37_1kg.AlignmentSummaryMetrics                         \
   R=/resources//hg19/indices/b37_1kg.fa                                     \
   VALIDATION_STRINGENCY=LENIENT                                             \


java -jar /tools/picard-tools-1.32/CollectGcBiasMetrics.jar \
   R=/resources//hg19/indices/b37_1kg.fa                    \
   I=/output/filename.b37_1kg.sorted.bam                     \
   O=/output/filename.b37_1kg.GcBiasMetrics                  \
   CHART=/output/filename.b37_1kg.GcBiasMetrics.pdf          \
   VALIDATION_STRINGENCY=LENIENT                            \


java -jar /tools/picard-tools-1.32/CollectInsertSizeMetrics.jar \
   I=/output/filename.b37_1kg.sorted.bam                         \
   O=/output/filename.b37_1kg.CollectInsertSizeMetrics           \
   H=/output/filename.b37_1kg.CollectInsertSizeMetrics.pdf       \
   VALIDATION_STRINGENCY=LENIENT                                \


java -jar /tools/picard-tools-1.32/MeanQualityByCycle.jar \
   I=/output/filename.b37_1kg.sorted.bam                   \
   O=/output/filename.b37_1kg.MeanQualityByCycle           \
   CHART=/output/filename.b37_1kg.MeanQualityByCycle.pdf   \
   VALIDATION_STRINGENCY=LENIENT                          \


java  -jar  /tools/picard-tools-1.32/QualityScoreDistribution.jar       \
   I=/output/filename.b37_1kg.sorted.bam                                 \
   O=/output/filename.b37_1kg.[wiki:QualityScoreDistribution]            \
   CHART=/output/filename.b37_1kg.[wiki:QualityScoreDistribution].pdf    \
   VALIDATION_STRINGENCY=LENIENT                                        \


java -jar /tools/picard-tools-1.32/BamIndexStats.jar \
   INPUT=/output/filename.b37_1kg.sorted.bam          \
   VALIDATION_STRINGENCY=LENIENT                     \


java -jar -Xmx3g /tools/picard-tools-1.32/CalculateHsMetricsWholeGenome.jar \
   INPUT=/output/filename.b37_1kg.sorted.bam                                 \
   OUTPUT=/output/filename.b37_1kg.HsMetrics                                 \
   BAIT_INTERVALS=/resources//hg19/intervals/GoNL.interval_list             \
   TARGET_INTERVALS=/resources//hg19/intervals/GoNL.interval_list           \
   VALIDATION_STRINGENCY=LENIENT                                            \

Split BAMS by Chromosome

BAMs are split into chromosomes BAMs. These files move on to variant calling.

Compress BAMS with ReduceRead

Variant Calling


java -jar GenomeAnalysisTK.jar \
  -R resources/Homo_sapiens_assembly18.fasta \
  -T UnifiedGenotyper \
  -I sample1.bam [-I sample2.bam ...] \
  --dbsnp dbSNP.vcf \
  -o snps.raw.vcf \
  -stand_call_conf [50.0] \
  -stand_emit_conf 10.0 \
  -dcov [50 for 4x, 200 for >30x WGS or Whole exome] \
  [-L targets.interval_list]


java -Xmx4g -jar GenomeAnalysisTK.jar \
  -T VariantRecalibrator \
  -R reference/human_g1k_v37.fasta \
  -input NA12878.HiSeq.WGS.bwa.cleaned.raw.subset.b37.vcf \
  -resource:hapmap,known=false,training=true,truth=true,prior=15.0 hapmap_3.3.b37.sites.vcf \
  -resource:omni,known=false,training=true,truth=false,prior=12.0 1000G_omni2.5.b37.sites.vcf \
  -resource:dbsnp,known=true,training=false,truth=false,prior=6.0 dbsnp_135.b37.vcf \
  -an QD -an HaplotypeScore -an MQRankSum -an ReadPosRankSum -an FS -an MQ -an InbreedingCoeff \
  -mode SNP \
  -recalFile path/to/output.recal \
  -tranchesFile path/to/output.tranches \
  -rscriptFile path/to/output.plots.R


java -Xmx3g -jar GenomeAnalysisTK.jar \
  -T ApplyRecalibration \
  -R reference/human_g1k_v37.fasta \
  -input NA12878.HiSeq.WGS.bwa.cleaned.raw.subset.b37.vcf \
  --ts_filter_level 99.0 \
  -tranchesFile path/to/output.tranches \
  -recalFile path/to/output.recal \
  -mode SNP \
  -o path/to/output.recalibrated.filtered.vcf

Variant File Analyses

Family Based Analyses



Population Based Analyses



Phenotype Based Analyses