Pipeline

Germline SNP and short indel pipelines

QC procedure

The advance in NGS technologies has greatly improved our ability to detect genomic variants for biomedical research. One of the major challenges is quality control of the sequencing data. CBM team of VinBigData created the quality control procedures at three different stages of sequencing data:

QC level 0 on non-alignment files (raw and clean fastq files) to generate non-alignment based metrics

QC level 1 on alignment data (bam file) to generate alignment metrics

QC level 2 on variant calling data (vcf file) for the assessment after bioinformatic analysis

QC level 0

Raw fastq

Object: conducting quality control of raw data from Novaseq sequencing

Tools: FastQC v0.11.7

Input: fastq files of read 1 and read 2

Output: FastQC report include the information about:

Basic statistics: File name, number of reads, GC content

Per base sequence quality, Per tile sequence quality, Per sequence quality scores

Per base sequence content, Per sequence GC content, Per base N content

Sequence length distribution, sequence duplication levels, Overrepresented sequences, adapter content

Clean fastq

Object: conducting quality control of data after 5 cleaning processes:

Trimming adapter, trimming polyG and base 151 of reads

Filter reads whose low quality bases (<=5) is over 50%

Filter reads whose unknown bases (base N) ratio is over more than 10%

Tools: FastQC v0.11.7, fastp v0.20 report of cleaning process

Input: fastq files of read 1 and read 2 after cleaning process

Output: FastQC report include the information about:

Basic statistics: File name, number of reads, GC content

Per base sequence quality, Per tile sequence quality, Per sequence quality scores

Per base sequence content, Per sequence GC content, Per base N content

Sequence length distribution, sequence duplication levels, Overrepresented sequences, adapter content

Calculate percentage of clean data and quality improvement of data after cleaning

QC level 1

Object: providing additional insights into sample sequencing quality and identify bad samples that pass the QC level 0 checks. The QC level 1 implements quality check on data after mapping to reference genome (version hg38).

Tools: Picard v2.20 CollectRawWGSMetrics and Dragen mapping metrics

Input: bam file after mapping to a refence genome by dragen mapping or bwa mem.

Output: alignment metrics including the information about:

Depth = mean coverage

Coverage at 4x, 15x achieved the expectation of 95% and 90% respectively.

Coverage over reference genome (expected 98%)

Percentage of reference genome covered at the sequencing at that depth

QC level 2

Object: quality control of variant calling file to minimize the rate of false-positive Variant calling step.

Tools: Dragen variant calling report and GATK v4.1tools: VariantFiltration, VariantRecalibrator, ApplyVQSR, ValidateVariants, CollectVariantCallingMetrics.

Input: vcf file after performing variant calling by different tools: HaplotypeCaller by GATK v4.0, DeepVariant v0.7.1, CNNScoreVariant by GATK v4.1

Output: variant calling metrics including the information about:

The total of variants: the number of SNP, MNP, bi-allelic, multi-allelic SNV

The Ti/Tv ratio: the number of transition SNP divided by the number of transversion SNP (For human genome data, the Ti/Tv ratio is around 3.0 for exon SNP and 2.0 elsewhere – exclude mtDNA).

The other criteria of each site such as DP, GQ, MQ, VQSLOD.

SV pipelines