TNseq

Matching the results of a GATK somatic pipeline with Sentieon

---

Introduction

This document describes the capabilities of Sentieon TNseq pipelines matching different versions of GATK somatic variant calling pipelines. If you have any questions, please visit https://www.sentieon.com/support or contact the technical support at Sentieon Inc. at support@sentieon.com.

Input files

Three somatic samples were processed. (1) Targeted sequencing data obtained from BaseSpace. (2) A formalin-fixed paraffin-embedded (FFPE) sample from the SEQC-II consortium. (3) A paired tumor-normal WGS sample obtained from BaseSpace.

Sample Type	Sample Source	Sample Identifier	Tumor FASTQ	Normal FASTQ
Amplicon	BaseSpace Project - NextSeq 550: Comprehensive v3 DNA Panel (Coriell and Horizon Samples)	HD200-10ng-E05C-H01	HD200-10ng-E05C-H01_S43_L001_R[12]_001.fastq.gz, HD200-10ng-E05C-H01_S43_L001_R[12]_001.fastq.gz, HD200-10ng-E05C-H01_S43_L001_R[12]_001.fastq.gz, HD200-10ng-E05C-H01_S43_L001_R[12]_001.fastq.gz	N/A
FFPE	Sequence Read Archive - SRR7890933 and SRR7890951	FFX_IL_T_24h_1/FFX_IL_N_24h_2	SRR7890933_[12].fastq.gz	SRR7890951_[12].fastq.gz
WGS	BaseSpace Project - HiSeq 2000: Tumor Normal WGS (HCC1187 & HCC2218)	HCC1187	HCC1187C_S1_L001_R[12]_001.fastq.gz	HCC1187BL_S1_L001_R[12]_001.fastq.gz

Reference files

The hs37d5 reference genome can be downloaded from ftp://ftp-trace.ncbi.nih.gov/1000genomes/ftp/technical/reference/phase2_reference_assembly_sequence//hs37d5.fa.gz. Other bundle files were downloaded from gs://gatk-best-practices/somatic-b37/.

Argument	Input file
fasta	hs37d5.fa
germline	small_exac_common_3.vcf

Preprocessing

All input fastq files were preprocessed through alignment, duplicate marking and BQSR using the Sentieon DNAseq pipeline. BQSR was applied to the aligned reads using the Sentieon ReadWriter for compatibility with the GATK tools.

Evaluation

The som.py tool from the hap.py package (https://github.com/Illumina/hap.py) was used to compare the Sentieon and GATK VCFs, with the GATK vcf as the "truth" VCF. The following command was used for the comparison:

python happy/bin/som.py \
  -o output/eval \
  -r $fasta \
  -N \
  $truth_vcf \
  $calls_vcf \
  --no-fixchr-truth \
  --no-fixchr-query

Somatic variant calling

In the commands below, the input alignment files after preprocessing are the tumor_bam and normal_bam for the tumor and normal preprocessed alignment files, respectively. Arguments and commands used only with the paired normal file are marked with square brackets, [ and ]. The GATK commands were parallelized across chromosomes and merged, but the parallelization and merging commands are omitted below for simplicity.

By default, the GATK will downsample reads during variant calling. To reduce the effect of downsampling, MuTect2 in the GATK3 was run with --maxReadsInRegionPerSample 1000000 --minReadsPerAlignmentStart 1000000 -dt NONE and Mutect2 in the GATK4 was run with --max-reads-per-alignment-start 0. Downsampling arguments are omitted in the command line below.

TNhaplotyper and MuTect2 version 3.8.1

Commands

GATK3 variant calling

java -jar GenomeAnalysisTK.jar \
  -T MuTect2 \
  -R $fasta \
  -I:tumor $tumor_bam \
  [-I:normal $normal_bam] \
  -o output.vcf.gz

Sentieon variant calling

sentieon driver \
  -r $fasta \
  -i $tumor_bam \
  [-i $normal_bam] \
  --algo TNhaplotyper \
  --tumor_sample $tumor_sample \
  [--normal_sample $normal_sample] \
  output.vcf.gz

Results

Sample Type	Variant Type	Matched	Sentieon Additional	Sentieon Missed	Sensitivity	Precision
Amplicon	SNVs	625	0	2	0.9968	1.0
Amplicon	INDELs	128	0	0	1.0	1.0
FFPE	SNVs	2027	0	0	1.0	1.0
FFPE	INDELs	250	0	0	1.0	1.0
WGS	SNVs	10950	0	0	1.0	1.0
WGS	INDELs	950	2	0	1.0	0.9979

Runtime

Samples were processed on a 32 virtual core Intel Xeon(R) CPU v3 machine (2.40GHz) with hyperthreading and 64GB memory. The Sentieon tools were parallelized using their built-in multi-threading while GATK jobs were parallelized across chromosomes and merged. Due to memory limitations, 8 GATK jobs were run in parallel.

Stage	Sentieon - FFPE	GATK - FFPE	Sentieon - WGS	GATK - WGS	Sentieon - Amplicion	GATK - Amplicon
Variant calling	03:38:07	36:56:53	00:09:10	01:54:48	00:00:47	00:08:56
Sentieon Speedup	10.2x	--	12.5x	--	11.4x	--

TNhaplotyper2 and Mutect2 version 4.1.9.0

Commands

GATK4 variant calling

gatk Mutect2 \
  -R $fasta \
  -I $tumor_bam \
  [-I $normal_bam] \
  -tumor $tumor_sample \
  [-normal $normal_sample] \
  --germline-resource $germline \
  -O tmp.vcf.gz \
  --f1r2-tar-gz f1r2.tar.gz

gatk GetPileupSummaries \
  -R $fasta \
  -I $tumor_bam \
  -V $germline \
  -O tumor_pileups.tsv

[gatk GetPileupSummaries \
  -R $fasta \
  -I $normal_bam \
  -V $germline \
  -O normal_pileups.tsv]

gatk LearnReadOrientationModel \
  -I f1r2.tar.gz \
  -O artifact-priors.tar.gz

gatk CalculateContamination \
  -I tumor_pileups.tsv \
  [-I normal_pileups.tsv] \
  -O contamination.table \
  --tumor-segmentation segments.table

gatk FilterMutectCalls \
  -V tmp.vcf.gz \
  -R $fasta \
  -O output.vcf.gz \
  --contamination-table contamination.table \
  --tumor-segmentation segments.table \
  --ob-priors artifact-priors.tar.gz \
  -stats tmp.vcf.gz.stats \
  --filtering-stats output.vcf.gz.stats

Sentieon variant calling

sentieon driver \
  -r $fasta \
  -i $tumor_bam \
  -i $normal_bam \
  --algo TNhaplotyper2 \
    --tumor_sample $tumor_sample \
    [--normal_sample $normal_sample] \
    --germline_vcf $germline \
    tmp.vcf.gz \
  --algo OrientationBias \
    --tumor_sample $tumor_sample \
    tmp.priors \
  --algo ContaminationModel \
    --tumor_sample $tumor_sample \
    [--normal_sample $normal_sample] \
    -v $germline \
    --tumor_segments tmp.segments \
    tmp.contam.table

sentieon driver \
  -r $fasta \
  --algo TNfilter \
  -v tmp.vcf.gz \
  --tumor_sample $tumor_sample \
  [--normal_sample $normal_sample] \
  --contamination tmp.contam.table \
  --tumor_segments tmp.segments \
  --orientation_priors tmp.priors \
  output.vcf.gz

Results

Sample Type	Variant Type	Matched	Sentieon Additional	Sentieon Missed	Sensitivity	Precision
Amplicon	SNVs	7758	24	48	0.9939	0.9969
Amplicon	INDELs	718	5	3	0.9958	0.9931
Amplicon	MNPs	168	0	0	1.0	1.0
FFPE	SNVs	4322	3	31	0.9929	0.9993
FFPE	INDELs	375	0	1	0.9973	1.0
FFPE	MNPs	191	0	2	0.9896	1.0
WGS	SNVs	11656	2	6	0.9995	0.9998
WGS	INDELs	1299	4	1	0.9992	0.9969
WGS	MNPs	299	0	0	1.0	1.0

Runtime

Samples were processed on a 32 virtual core Intel Xeon(R) CPU v3 machine (2.40GHz) with hyperthreading and 64GB memory. The Sentieon tools were parallelized using their built-in multi-threading while GATK Mutect2 jobs were parallelized across chromosomes and merged. Due to memory limitations, up to 16 GATK jobs were run in parallel. Some GATK tasks were run concurrently, so the GATK's overall time is less than the sum of the task time.

Stage	Sentieon - FFPE	GATK - FFPE	Sentieon - WGS	GATK - WGS	Sentieon - Amplicon	GATK - Amplicon
Variant calling	00:22:19	02:55:32	00:33:26	01:49:20	00:00:23	00:03:52
GATK Pileups	--	02:13:29	--	01:34:36	--	00:01:00
GATK Contamination	--	00:00:12	--	00:00:10	--	00:00:04
GATK LearnReadOri	--	00:14:37	--	00:09:36	--	00:00:28
Variant filtering	00:00:11	00:02:33	00:00:10	00:01:44	00:00:04	00:00:40
Overall wall time	00:22:30	03:12:42	00:33:36	02:00:40	00:00:27	00:05:00
Sentieon Speedup	8.6x	--	3.6x	--	11.1x	--

GATK4 tasks were run concurrently. The plot only includes the "rate-limiting" GATK4 tasks that sum to the overall wall time.