TeraPCA: a fast and scalable software package to study genetic variation in tera-scale genotypes

Area of Science:

• Human genetics
• Population structure analysis
• Bioinformatics

Background:

• Principal Component Analysis (PCA) is crucial for understanding population structure in human genetics.
• Increasingly large genetic datasets challenge traditional in-memory PCA methods.
• Out-of-core implementations are necessary for analyzing massive genomic data.

Purpose of the Study:

• To present TeraPCA, an efficient C++ implementation for large-scale Principal Component Analysis.
• To enable PCA on datasets that exceed available system memory.
• To provide a practical tool for human population genetics research.

Main Methods:

• Implemented Randomized Subspace Iteration for PCA.
• Developed TeraPCA as a C++ software package.
• Designed for both in-core and out-of-core computation on commodity hardware.

Main Results:

• TeraPCA successfully performs PCA on large-scale datasets using minimal RAM.
• Achieved accurate computation of 10 leading principal components for a dataset of one million individuals and one million markers in under 5 hours.
• Demonstrated competitive speed and accuracy compared to existing state-of-the-art PCA software.

Conclusions:

• TeraPCA offers a fast and accurate solution for large-scale PCA in human genetics.
• The software is accessible and runs effectively on standard hardware.
• Enables advanced population structure studies with massive genomic datasets.