GRAF

Seven Bridges GRAF™ is a pan-genome bioinformatics suite designed for secondary analysis of next-generation sequencing (NGS) data using graph-based genome references. Unlike traditional linear haploid reference genomes, GRAF™ employs directed acyclic graph structures that naturally represent the full spectrum of human genetic diversity — including SNPs, INDELs, and structural variants — enabling variant calling with superior accuracy without compromising speed or cost. The suite is built for researchers and organisations conducting large-scale population genomics, rare disease studies, and personalised genome analysis.

GRAF™ uses standard genomic data formats including FASTA, VCF, FASTQ, BAM, CRAM, and BED, ensuring compatibility with existing bioinformatics workflows. All workflows are specified in CWL 1.0, guaranteeing portability across diverse compute environments. The suite supports large human population studies, precise individual genome analysis using personalised and family genome graphs for accurate de novo mutation detection, and rare disease studies using curated disease-associated mutation graphs.

Workflows

• GRAF™ Germline Variant Detection Workflow – Uses a Pan-Genome graph containing genetic information from populations around the world. Accepts input reads in FASTQ, CRAM, or BAM format, produces graph-based alignments in BAM or CRAM format, and outputs variant calls in VCF format.
• GRAF™ Extended Germline Variant Detection Workflow – Makes use of an ancestry-aware, population-specific genome graph to increase sensitivity towards the represented population. Includes a downstream supplementary step that merges and annotates all variants, outputting a single multi-sample VCF file. Particularly suited for targeted population studies.

Graph References

• GRAF™ Pan Genome Reference – Contains both small variants (SNPs and INDELs up to several dozen base-pairs) and larger structural variants that are typically difficult to identify from short-read sequencing data, reducing spurious variant calls.
• GRAF™ Admixed American Reference – Curated from approximately 8,000 samples from various public databases covering the whole genome, augmented with population-specific structural variants obtained from PacBio HiFi reads.
• GRAF™ African Reference – Curated from more than 20,000 samples from various public databases, augmented with population-specific structural variants from PacBio HiFi reads.
• GRAF™ East Asian Reference – Curated from more than 3,000 samples from various public databases, augmented with population-specific structural variants from PacBio HiFi reads.
• GRAF™ European Reference – Curated from more than 34,000 samples from various public databases, augmented with population-specific structural variants from PacBio HiFi reads.
• GRAF™ South Asian Reference – Curated from approximately 3,000 samples from various public databases, augmented with population-specific structural variants from PacBio HiFi reads.

Core Tools

• GRAF™ Aligner – Maps sample reads against a Genome Graph Reference, implicitly considering many alternate haplotypes at each locus to minimise reference bias. Incorporates variant data from sources including the 1,000 Genomes Project and the Simons Genome Diversity Project.
• GRAF™ Variant Caller – Enables integrated calling of SNPs and INDELs as well as structural variants present in the Genome Graph, supporting direct genotyping against the graph for known variants.
• GRAF™ Stats – Calculates graph-specific alignment metrics and measures the utilisation of the graph reference in read alignment, alongside standard metrics such as coverage, unmapped and multi-mapped read counts, and improper alignments.
• GRAF™ Genome Viewer – A specialised genome viewer capable of visualising graph references and alignments made against the edges of a graph reference.

Graph Construction Service

• GRAF™ Construction Service – Provides bioinformatics methods and services for building representative and accurate genome graph references tailored to specific use cases, ranging from pan-genome and population studies to personalised medicine and disease-related applications.

How Graph-Based Analysis Improves Variant Calling

• The GRAF™ Pan Genome Reference organises genomic data into an edge-based sequence variation graph, where edges are the primary data carriers and alternative haplotypes are represented as different paths through the graph.
• The linear reference assembly forms the graph backbone, with additional variants added as new edges; longer haplotypes are obtained by following a path through the graph and concatenating subsequences from visited edges.
• Graph structures can learn from every newly sequenced individual, meaning the reference improves with each additional genome added, with only minimal increases in file size — enabling population-scale analysis.
• Reads partially aligned to a known variant edge allow for better adaptive selection of the reassembly window, overcoming the limitations of fixed-size reassembly regions used in conventional linear variant callers.
• For variants already present in the graph, sequence and location information is known, enabling direct genotyping against the graph rather than relying solely on local reassembly.
• By preserving all variant information across populations, GRAF™ facilitates more accurate alignment and variant calling compared to linear reference-based alternatives.

Seven Bridges GRAF™ is suitable for BioPharma, healthcare, government, and academic organisations engaged in large-scale genomics research. Its use of open standard formats and CWL 1.0 workflow specifications ensures that it integrates readily into existing NGS pipelines and can be deployed across a variety of compute environments.