NovoGraph: Human genome graph construction from multiple long-read de novo assemblies

Area of Science:

• Genomics
• Bioinformatics

Background:

• Genome graphs enhance analysis of high-throughput human sequencing data, particularly for variable genomic regions.
• Current methods often rely on short-read sequencing and variant call sets.
• Long-read sequencing advances necessitate genome graph construction directly from assembled genomes.

Purpose of the Study:

• To present NovoGraph, a novel method for constructing human genome graphs directly from de novo assemblies.
• To enable the representation of a wider spectrum of genetic variation, including structural variants and divergent haplotypes, accessible via long-read sequencing.
• To output genome graphs in a format compatible with existing bioinformatics tools.

Main Methods:

• NovoGraph performs a genome-wide multiple sequence alignment of input contigs from de novo assemblies.
• It merges homologous and sequence-identical positions to create the genome graph.
• Output graphs are generated in VCF format for use with third-party genome graph toolkits.

Main Results:

• A human genome graph was constructed from seven diverse human genomes, featuring over 23 million variant sites and 30 million variant alleles.
• Mapping sequencing reads against the NovoGraph-generated graph reduced the average mismatch rate by approximately 0.2% for sample NA12878.
• A slight increase in unmapped reads was observed during initial evaluations.

Conclusions:

• NovoGraph provides a method for direct genome graph construction from de novo assemblies, expanding genomic variation representation.
• This approach facilitates improved analysis of complex genomic regions using long-read sequencing data.
• The generated graphs are compatible with existing tools, promoting wider adoption in genomic research.