Pangenome-based genome inference
Authors/Creators
- 1. Institute for Medical Biometry and Bioinformatics, Heinrich Heine University Düsseldorf, Düsseldorf, Germany
- 2. New York Genome Center, New York, New York, USA
- 3. European Molecular Biology Laboratory (EMBL), Genome Biology Unit, Heidelberg, Germany; European Molecular Biology Laboratory (EMBL), GeneCore, Heidelberg, Germany
- 4. Department of Genome Sciences, University of Washington School of Medicine, Seattle, Washington, USA
- 5. Institute of Medical Microbiology and Hospital Hygiene, Heinrich Heine University Düsseldorf, Düsseldorf, Germany
- 6. European Molecular Biology Laboratory (EMBL), Genome Biology Unit, Heidelberg, Germany
- 7. Department of Genome Sciences, University of Washington School of Medicine, Seattle, Washington, USA; Howard Hughes Medical Institute, University of Washington, Seattle, Washington, USA
- 8. Institute of Medical Microbiology and Hospital Hygiene, Heinrich Heine University Düsseldorf, Düsseldorf
Description
Variant calls (small panel + extended panel) produced from haplotype-resolved assemblies ("reference-panel.tar.gz"), PanGenie genotypes for 100 randomly selected 1000 Genomes samples ("cohort-genotypes.tar.gz"), results of a "leave-one-out experiment" for sample HG00731 ("leave-one-out-results.tar.gz").
Abstract:
Typical analysis workflows map reads to a reference genome in order to genotype genetic variants. Generating such alignments introduces reference biases and comes with substantial computational burden.
In contrast, recent k-mer based genotypers are fast, but struggle in repetitive or duplicated genomic regions.
We propose a novel algorithm, PanGenie, that leverages a pangenome reference in conjunction with k-mer counts from short-read sequencing data to genotype a wide spectrum of genetic variation.
Compared to mapping-based approaches, our approach is more than 4x faster at 30x coverage and reaches significantly better genotype concordances for almost all variant types and coverages tested. Improvements are especially pronounced for large insertions (>50bp), where we are able to genotype >99.9% of all tested variants with over 90% accuracy at 30x short-read coverage, where the best competing tools either typed less than 60% of variants or reached accuracies below 70%.