Pangenome-based human genome analysis improves trait association and genomic prediction

Area of Science:

• Genomics
• Human Genetics
• Bioinformatics

Background:

• The single linear reference genome limits comprehensive human genetic analysis.
• Pangenome reference graphs offer a more inclusive representation of human genetic diversity.
• Assessing the utility of pangenome approaches for trait mapping remains a key challenge.

Purpose of the Study:

• To evaluate the impact of pangenome-based methods on trait mapping, specifically using gene expression variation as a model.
• To develop and validate a novel graph-based method for associating sequence variation with traits.
• To quantify improvements in detecting expression quantitative trait loci (eQTLs) and enhancing genomic prediction.

Main Methods:

• Developed EdgeDepth, a graph-based method for trait association using short-read sequencing data.
• Utilized 430 samples with deep RNA-sequencing data for performance evaluation.
• Compared pangenomic methods against the 1000 Genomes Project reference callset.

Main Results:

• Pangenomic methods detected eQTLs involving multiallelic indels and structural variants, increasing power for a subset of genes.
• 812 genes showed ≥20% improvement in statistical significance, with 185 showing ≥50% improvement.
• Identified GBAP1 pseudogene copy number variation as a potential causal factor in Crohn's disease.
• Improved gene expression prediction models, increasing median variance explained from 10.1% to 12.5%.

Conclusions:

• Pangenome-based approaches significantly enhance trait association studies by capturing complex genetic variations.
• These methods improve the power to detect eQTLs and boost genomic prediction accuracy.
• Integration of pangenomic methods is crucial for advancing human genetic studies and understanding disease etiology.