Tools for comprehensive variant characterization using the pangenome

The U.S. National Human Genome Research Institute (NHGRI) awarded a $1,696,968 Cooperative Agreement grant (CFDA 93.172 - Human Genome Research) to the University of California, Santa Cruz to develop novel computational methods and software tools for analyzing pangenomic data, with a focus on improving the accuracy and efficiency of variant calling and genotyping, particularly for complex structural variants.

The grant, titled "Tools for Comprehensive Variant Characterization Using the Pangenome", includes five specific aims: 1) Creating a pangenome mapper supporting long-reads, 2) Developing personalized pangenomes, 3) Establishing a pangenome variant calling and imputation method, 4) Genotyping complex structural variants involving protein-coding genes, and 5) Expanding the definition of reportable variation through RGFA-based variant calling pipelines.

Two sub-awards were issued under this grant - one to the Heinrich-Heine-Universität Düsseldorf to develop a targeted genotyping method for complex structural variants, and another to the Dana-Farber Cancer Institute to work on pangenome construction, gene-level structural variation identification, and haplotype-specific genotyping.

Abstract

This grant proposal outlines a comprehensive plan to develop novel computational methods and software tools for analyzing pangenomic data, with a focus on improving the accuracy and efficiency of variant calling and genotyping, particularly for complex structural variants (SVS). The proposal is divided into five specific aims: aim 1: create a pangenome mapper supporting long-reads, which will enable accurate and efficient mapping of long-range sequencing data to pangenome references. Aim 2: develop personalized pangenomes, which involves rapid and efficient construction of a subset of a larger graph based on an input sample's k-mers. This approach will tailor the pangenome for specific analysis and so lead to improved performance in downstream analysis. Aim 3: create a pangenome variant calling and imputation method for unified genome inference, which will combine imputation with read-based genotyping using machine learning to infer a more complete representation of variation, including both small variants and SVS. Aim 4: genotyping complex SVS involving protein-coding genes, which will involve identifying long segmental duplications, grouping haplotypes, and developing targeted genotyping methods for long and short reads. Aim 5: develop mature rgfa based variant calling for reporting both SV and small variants within polymorphic sequence, which will expand the current definition of reportable variation and provide pipelines that can report tens of thousands of additional variations per sample. The proposal highlights the need for better computational tools for pangenome analysis, especially for complex SVS, and outlines a comprehensive plan to address these challenges. The proposed software tools and methods will enable researchers to analyze pangenomic data more effectively and efficiently, leading to new insights into genetic variation and its role in disease and other biological processes.