GGCPan: A graph-modeled gastric cancer pangenome

Introduction

We introduced a graph-based pangenome called GGCPan as a reference for gastric cancer, and systematically compared the results with those traditional genomics studies using the human reference genome or a linear pangenome as the references. This is our construction method and the analysis pipeline.

GGCPan Construction

• step1：Align assembled contigs(>500bp) to GRCh38 using minimap2.
• step2：Detect variants using paftools.js and filter the small variants and variants with quality <= 60.
• step3：Embed variants to GRCh38 using vg toolkit.

step1,2,3 are included in the following construct.sh file, which can be run to generate the graph-modeled pangenome.

bash construct.sh

Alignment

Reads aligned to linear references using BWA MEM and GATK pipeline.

• We firstly align the raw reads to references with BWA MEM,then mark duplications and adjust the base quality with GATK BQSR.
• The codes are stored in Alignment/gatk.slurm.

Reads aligened to graph-modeled pangenome using vg giraffe.

• The raw reads are aligned to graph-modeled reference with vg giraffe.
• To detect snps and indels from graph-modeled pangenome,we convert graph alignment format(.gam) to linear alignment format(.bam).
• The codes are stored in Alignment/graph.alignment.sh.

Variant Calling

SNPs and Indels

• We used GATK Mutect2 to detect SNPs and Indels.
• The coded are stored in VariantCalling/mutect2.sh.

Structural vatiants

• We used Manta, Delly,Svaba and Survivor to detect SVs based on linear references. We used vg call to detect SVs based on graph-modeled references.
• The codes are stored in VariantCalling/linear.variantDetection.sh and VariantCalling/graph.variantDetection.sh.

Structural Variant Results

• somatic.vcf is the somatic structural variants generated by our graph modeled pangenome.