FaST-LMM

FaST-LMM performs linear mixed-model association testing for genome-wide association studies (GWAS) by factoring and spectrally transforming models to reduce runtime and memory and enable scalable analysis of large cohorts.

Key Features:

• Scalability: Scales linearly with cohort size and can analyze datasets of up to 120,000 individuals within hours.
• Performance Efficiency: Runs an order of magnitude faster than other efficient algorithms on the Wellcome Trust dataset (15,000 individuals).
• Memory Optimization: Maintains linear memory scaling to manage datasets beyond typical limits where traditional algorithms struggle (≈20,000 individuals).

Scientific Applications:

• Genome-wide association studies (GWAS): Enables large-cohort association testing to identify genetic variants linked to complex traits and diseases.
• Prediction tasks: Supports genetic prediction analyses such as polygenic scoring on large datasets.
• Heritability estimation: Facilitates estimation of trait heritability using linear mixed-model frameworks.

Methodology:

The method factors and spectrally transforms linear mixed models to reduce runtime and memory requirements, enabling linear scaling with cohort size.