GROMTools: scalable individual-level GReX imputation for mega-biobank-scale cohorts

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• The increasing size of human biobanks and complexity of transcriptomic imputation models present computational challenges.
• Existing imputation tools are memory-intensive and ill-suited for cloud or shared computing environments, increasing costs and limiting throughput.
• Scalable and efficient imputation of individual-level genetically regulated gene expression (GReX) is crucial for large-scale genetic studies.

Purpose of the Study:

• To develop a computationally efficient engine for individual-level GReX imputation.
• To address the limitations of existing tools in handling large biobank-scale datasets.
• To provide a cost-effective solution for routine GReX imputation.

Main Methods:

• Developed GROMTools, an open-source C++ engine with an R interface.
• GROMTools utilizes sparse prediction weights and streams PLINK2 genotypes.
• Outputs compact binary files for efficient storage and retrieval.

Main Results:

• GROMTools achieved near-identical imputation accuracy to existing methods (Pearson correlation >0.999, RMSE <0.001) in UK Biobank benchmarks.
• Reduced CPU time by approximately 100-fold compared to existing tools.
• Decreased peak memory usage by approximately 33-fold.
• Demonstrated practical and cost-efficient biobank-scale GReX imputation on standard CPU infrastructure.

Conclusions:

• GROMTools offers a scalable and efficient solution for individual-level GReX imputation.
• The tool's performance gains make large-scale transcriptomic imputation feasible and cost-effective.
• GROMTools facilitates broader application of GReX imputation in large biobank studies.