LDAK-KVIK performs fast and powerful mixed-model association analysis of quantitative and binary phenotypes

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Mixed-model association analysis (MMAA) is crucial for genome-wide association studies (GWAS).
• Existing MMAA tools face challenges with long runtimes and high memory usage.
• Efficient MMAA is needed for large-scale genetic datasets.

Purpose of the Study:

• Introduce LDAK-KVIK, a computationally efficient MMAA tool.
• Evaluate LDAK-KVIK's performance for quantitative and binary phenotypes.
• Compare LDAK-KVIK's power and accuracy against existing MMAA methods.

Main Methods:

• Developed LDAK-KVIK for MMAA of quantitative and binary traits.
• Assessed computational efficiency (CPU hours, memory) on large datasets (350,000 individuals).
• Validated test statistic calibration using simulated phenotypes (homogeneous and heterogeneous).

Main Results:

• LDAK-KVIK requires <10 CPU hours and <5 Gb memory for genome-wide analysis of 350,000 individuals.
• Achieved well-calibrated test statistics on simulated data.
• Demonstrated superior power in identifying genome-wide significant loci and genes in UK Biobank data compared to classical linear regression, BOLT-LMM, and REGENIE.
• Produced state-of-the-art polygenic scores.

Conclusions:

• LDAK-KVIK offers a computationally efficient and powerful alternative for MMAA.
• The tool enhances the discovery of genetic associations and improves polygenic score prediction.
• LDAK-KVIK is suitable for large-scale genetic studies with diverse phenotypes.