Quality control of germline variants calling results using a Chinese Quartet family

Author： Ren Luyao, Chen Haonan

E-mail：18110700050@fudan.edu.cn, haonanchen0815@163.com

Git: http://choppy.3steps.cn/chenhaonan/quartet_dna_quality_control_wes_big_pipeline.git

Last Updates: 2023/7/31

Install

open-choppy-env
choppy install chenhaonan/quartet_dna_quality_control_big_pipeline

Introduction

Chinese Quartet DNA reference materials

With the rapid development of sequencing technology and the dramatic decrease of sequencing costs, DNA sequencing has been widely used in scientific research, diagnosis of and treatment selection for human diseases. However, due to the lack of effective quality assessment and control of the high-throughput omics data generation and analysis processes, variants calling results are seriously inconsistent among different technical replicates, batches, laboratories, sequencing platforms, and analysis pipelines, resulting in irreproducible scientific results and conclusions, huge waste of resources, and even endangering the life and health of patients. Therefore, reference materials for quality control of the whole process from omics data generation to data analysis are urgently needed.

We first established genomic DNA reference materials from four immortalized B-lymphoblastoid cell lines of a Chinese Quartet family including parents and monozygotic twin daughters to make performance assessment of germline variants calling results. To establish small variant benchmark calls and regions, we generated whole-genome sequencing data in nine batches, with depth ranging from 30x to 60x, by employing PCR-free and PCR libraries on four popular short-read sequencing platforms (Illumina HiSeq XTen, Illumina NovaSeq, MGISEQ-2000, and DNBSEQ-T7) with three replicates at each batch, resulting in 108 libraries in total and 27 libraries for each Quartet DNA reference material. Then, we selected variants concordant in multiple call sets and in Mendelian consistency within Quartet family members as small variant benchmark calls, resulting in 4.2 million high-confidence variants (SNV and Indel) and 2.66 G high confidence genomic region, covering 87.8% of the human reference genome (GRCh38, chr1-22 and X). Two orthogonal technologies were used for verifying the high-confidence variants. The consistency rate with PMRA (Axiom Precision Medicine Research Array) was 99.6%, and 95.9% of high-confidence variants were validated by 10X Genomics whole-genome sequencing data. Genetic built-in truth of the Quartet family design is another kind of “truth” within the four Quartet samples. Apart from comparison with benchmark calls in the benchmark regions to identify false-positive and false-negative variants, pedigree information among the Quartet DNA reference materials, i.e., reproducibility rate of variants between the twins and Mendelian concordance rate among family members, are complementary approaches to comprehensively estimate genome-wide variants calling performance. Finally, we developed a whole-genome sequencing data quality assessment pipeline and demonstrated its utilities with two examples of using the Quartet reference materials and datasets to evaluate data generation performance in three sequencing labs and different data analysis pipelines.

Quality control pipeline for WGS

This Quartet quality control pipeline evaluate the performance of reads quality and variant calling quality. This pipeline accepts FASTQ format input files or VCF format input files. If the users input FASTQ files, this APP will output the results of pre-alignment quality control from FASTQ files, post-alignment quality control from BAM files and variant calling quality control from VCF files. GATK best practice pipelines (implemented by SENTIEON software) were used to map reads to the reference genome and call variants. If the users input VCF files, this APP will only output the results of variant calling quality control.

Quartet quality control analysis pipeline started from FASTQ files is implemented across seven main procedures:

• Pre-alignment QC of FASTQ files
• Genome alignment
• Post-alignment QC of BAM files
• Germline variant calling
• Variant calling QC depended on benchmark sets of VCF files
• Check Mendelian ingeritance states across four Quartet samples of every variants
• Variant calling QC depended on Quartet genetic relationship of VCF files

Quartet quality control analysis pipeline started from VCF files is implemented across three main procedures:

• Variant calling QC depended on benchmark sets of VCF files
• Check Mendelian ingeritance states across four Quartet samples of every variants
• Variant calling QC depended on Quartet genetic relationship of VCF files

Results generated from this APP can be visualized by Choppy report.

Data Processing Steps

1. Pre-alignment QC of FASTQ files

Fastqc v0.11.5

FastQC is used to investigate the quality of fastq files

fastqc -t <threads> -o <output_directory> <fastq_file>

Fastq Screen 0.12.0

Fastq Screen is used to inspect whether the library were contaminated. For example, we expected 99% reads aligned to human genome, 10% reads aligned to mouse genome, which is partly homologous to human genome. If too many reads are aligned to E.Coli or Yeast, libraries or cell lines are probably comtminated.

fastq_screen --aligner <aligner> --conf <config_file> --top <number_of_reads> --threads <threads> <fastq_file>

2. Genome alignment

Reads were mapped to the human reference genome GRCh38 using BWA-MEM.SAMTools is a tool used for SAM/BAM file conversion and BAM file sorting.

BWA-MEM:v0.7.17

# Mapping to reference genome, converting sam to bam, sorting bam file 
bwa mem -M -R "@RG\tID:${group}\tSM:${sample}\tPL:${pl}" -t $(nproc) -K 10000000 ${ref_dir}/${fasta} ${fastq_1} ${fastq_2} | samtools view -bS -@ $(nproc) - 
| samtools sort -@ $(nproc) -o ${user_define_name}_${project}_${sample}.sorted.bam -

SAMTools:v1.17

# Building an index for sorted bam file
samtools index -@ $(nproc) -o ${user_define_name}_${project}_${sample}.sorted.bam.bai ${user_define_name}_${project}_${sample}.sorted.bam

3. Post-alignment QC

Qualimap and Picard Tools are used to check the quality of BAM files. Deduplicated BAM files are used in this step.

Qualimap 2.0.0

# BAM QC by qualimap
qualimap bamqc -bam <bam_file> -outformat PDF:HTML -nt <threads> -outdir <output_directory> --java-mem-size=32G 

Picard:v3.0.0

# Remove duplicates
java -jar /usr/local/picard.jar  MarkDuplicates -I ${sorted_bam} -O ${sample}.sorted.deduped.bam -M ${sample}_dedup_metrics.txt --REMOVE_DUPLICATES
# Building an index for the sorted and deduplicated bam file			
samtools index -@ $(nproc) -o  ${sample}.sorted.deduped.bam.bai  ${sample}.sorted.deduped.bam

4. Germline variant calling

HaplotyperCaller implemented by Google DeepVariant is used to identify germline variants.

DeepVariant 1.5.0

# Calling variant
deepvariant/bin/run_deepvariant --model_type=WES --ref=${ref_dir}/${fasta} --reads=${recaled_bam} --output_vcf=${sample}_hc.vcf --num_shards=$(nproc)
# Building an index for the vcf file
gatk IndexFeatureFile -I ${sample}_hc.vcf -O ${sample}_hc.vcf.idx

5. Variants Calling QC

5.1 Performance assessment based on benchmark sets

Hap.py v0.3.9

Variants were compared with benchmark calls in benchmark regions.

hap.py <truth_vcf> <query_vcf> -f <bed_file> --threads <threads> -o <output_filename>

5.2 Performance assessment based on Quartet genetic built-in truth

VBT v1.1

We splited the Quartet family to two trios (F7, M8, D5 and F7, M8, D6) and then do the Mendelian analysis. A Quartet Mendelian concordant variant is the same between the twins (D5 and D6) , and follow the Mendelian concordant between parents (F7 and M8). Mendelian concordance rate is the Mendelian concordance variant divided by total detected variants in a Quartet family. Only variants on chr1-22,X are included in this analysis.

vbt mendelian -ref <fasta_file> -mother <family_merged_vcf> -father <family_merged_vcf> -child <family_merged_vcf> -pedigree <ped_file> -outDir <output_directory> -out-prefix <output_directory_prefix> --output-violation-regions -thread-count <threads>

Input files

choppy samples renluyao/quartet_dna_quality_control_wgs_big_pipeline-latest --output samples

Samples CSV file

1. Start from Fastq files

sample_id,project,fastq_1_D5,fastq_2_D5,fastq_1_D6,fastq_2_D6,fastq_1_F7,fastq_2_F7,fastq_1_M8,fastq_2_M8
# sample_id in choppy system
# project name
# oss path of D5 fastq read1 file
# oss path of D5 fastq read2 file
# oss path of D6 fastq read1 file
# oss path of D6 fastq read2 file
# oss path of F7 fastq read1 file
# oss path of F7 fastq read2 file
# oss path of M8 fastq read1 file
# oss path of M8 fastq read2 file

2. Start from VCF files

sample_id,project,vcf_D5,vcf_D6,vcf_F7,vcf_M8
# sample_id in choppy system
# project name
# oss path of D5 VCF file
# oss path of D6 VCF file
# oss path of F7 VCF file
# oss path of M8 VCF file

Output Files

1. extract_tables.wdl/extract_tables_vcf.wdl

(FASTQ) Pre-alignment QC: pre_alignment.txt

(FASTQ) Post-alignment QC: post_alignment.txt

(FASTQ/VCF) Variants calling QC: variants.calling.qc.txt

2. quartet_mendelian.wdl

(FASTQ/VCF) Variants calling QC: mendelian.txt

Output files format

1. pre_alignment.txt

Column name	Description
Sample	Sample name
%Dup	Percentage duplicate reads
%GC	Average GC percentage
Total Sequences (million)	Total sequences
%Human	Percentage of reads mapped to human genome
%EColi	Percentage of reads mapped to Ecoli
%Adapter	Percentage of reads mapped to adapter
%Vector	Percentage of reads mapped to vector
%rRNA	Percentage of reads mapped to rRNA
%Virus	Percentage of reads mapped to virus
%Yeast	Percentage of reads mapped to yeast
%Mitoch	Percentage of reads mapped to mitochondrion
%No hits	Percentage of reads not mapped to genomes mentioned above

2. post_alignment.txt

Column name	Description
Sample	Sample name
%Mapping	Percentage of mapped reads
%Mismatch Rate	Mapping error rate
Mendelian Insert Size	Median insert size（bp）
%Q20	Percentage of bases >Q20
%Q30	Percentage of bases >Q30
Mean Coverage	Mean deduped coverage
Median Coverage	Median deduped coverage
PCT_1X	Fraction of genome with at least 1x coverage
PCT_5X	Fraction of genome with at least 5x coverage
PCT_10X	Fraction of genome with at least 10x coverage
PCT_30X	Fraction of genome with at least 30x coverage

3. variants.calling.qc.txt

Column name	Description
Sample	Sample name
SNV number	Total SNV number (chr1-22,X)
INDEL number	Total INDEL number (chr1-22,X)
SNV query	SNV number in benchmark region
INDEL query	INDEL number in benchmark region
SNV TP	True positive SNV
INDEL TP	True positive INDEL
SNV FP	False positive SNV
INDEL FP	True positive INDEL
SNV FN	False negative SNV
INDEL FN	False negative INDEL
SNV precision	Precision of SNV calls when compared with benchmark calls in benchmark regions
INDEL precision	Precision of INDEL calls when compared with benchmark calls in benchmark regions
SNV recall	Recall of SNV calls when compared with benchmark calls in benchmark regions
INDEL recall	Recall of INDEL calls when compared with benchmark calls in benchmark regions
SNV F1	F1 score of SNV calls when compared with benchmark calls in benchmark regions
INDEL F1	F1 score of INDEL calls when compared with benchmark calls in benchmark regions

4 {project}.summary.txt

Column name	Description
Family	Family name defined by inputed project name
Reproducibility_D5_D6	Percentage of variants were shared by the twins (D5 and D6)
Mendelian_Concordance_Quartet	Percentage of variants were Mendelian concordance