Dataset Information

OCMA: Fast, Memory-Efficient Factorization of Prohibitively Large Relationship Matrices.

ABSTRACT: Matrices representing genetic relatedness among individuals (i.e., Genomic Relationship Matrices, GRMs) play a central role in genetic analysis. The eigen-decomposition of GRMs (or its alternative that generates fewer top singular values using genotype matrices) is a necessary step for many analyses including estimation of SNP-heritability, Principal Component Analysis (PCA), and genomic prediction. However, the GRMs and genotype matrices provided by modern biobanks are too large to be stored in active memory. To accommodate the current and future "bigger-data", we develop a disk-based tool, Out-of-Core Matrices Analyzer (OCMA), using state-of-the-art computational techniques that can nimbly perform eigen and Singular Value Decomposition (SVD) analyses. By integrating memory mapping (mmap) and the latest matrix factorization libraries, our tool is fast and memory-efficient. To demonstrate the impressive performance of OCMA, we test it on a personal computer. For full eigen-decomposition, it solves an ordinary GRM (N = 10,000) in 55 sec. For SVD, a commonly used faster alternative of full eigen-decomposition in genomic analyses, OCMA solves the top 200 singular values (SVs) in half an hour, top 2,000 SVs in 0.95 hr, and all 5,000 SVs in 1.77 hr based on a very large genotype matrix (N = 1,000,000, M = 5,000) on the same personal computer. OCMA also supports multi-threading when running in a desktop or HPC cluster. Our OCMA tool can thus alleviate the computing bottleneck of classical analyses on large genomic matrices, and make it possible to scale up current and emerging analytical methods to big genomics data using lightweight computing resources.

SUBMITTER: Xiong Z

PROVIDER: S-EPMC6325911 | biostudies-other | 2019 Jan

REPOSITORIES: biostudies-other

ACCESS DATA

Publications

OCMA: Fast, Memory-Efficient Factorization of Prohibitively Large Relationship Matrices.

Xiong Zhi Z Zhang Qingrun Q Platt Alexander A Liao Wenyuan W Shi Xinghua X de Los Campos Gustavo G Long Quan Q

G3 (Bethesda, Md.) 20190109 1

Matrices representing genetic relatedness among individuals (<i>i.e.</i>, Genomic Relationship Matrices, GRMs) play a central role in genetic analysis. The eigen-decomposition of GRMs (or its alternative that generates fewer top singular values using genotype matrices) is a necessary step for many analyses including estimation of SNP-heritability, Principal Component Analysis (PCA), and genomic prediction. However, the GRMs and genotype matrices provided by modern biobanks are too large to be st ...[more]

PMID: 30482799

Similar Datasets

Project description:Principal component analysis (PCA) is widely used for data reduction in group independent component analysis (ICA) of fMRI data. Commonly, group-level PCA of temporally concatenated datasets is computed prior to ICA of the group principal components. This work focuses on reducing very high dimensional temporally concatenated datasets into its group PCA space. Existing randomized PCA methods can determine the PCA subspace with minimal memory requirements and, thus, are ideal for solving large PCA problems. Since the number of dataloads is not typically optimized, we extend one of these methods to compute PCA of very large datasets with a minimal number of dataloads. This method is coined multi power iteration (MPOWIT). The key idea behind MPOWIT is to estimate a subspace larger than the desired one, while checking for convergence of only the smaller subset of interest. The number of iterations is reduced considerably (as well as the number of dataloads), accelerating convergence without loss of accuracy. More importantly, in the proposed implementation of MPOWIT, the memory required for successful recovery of the group principal components becomes independent of the number of subjects analyzed. Highly efficient subsampled eigenvalue decomposition techniques are also introduced, furnishing excellent PCA subspace approximations that can be used for intelligent initialization of randomized methods such as MPOWIT. Together, these developments enable efficient estimation of accurate principal components, as we illustrate by solving a 1600-subject group-level PCA of fMRI with standard acquisition parameters, on a regular desktop computer with only 4 GB RAM, in just a few hours. MPOWIT is also highly scalable and could realistically solve group-level PCA of fMRI on thousands of subjects, or more, using standard hardware, limited only by time, not memory. Also, the MPOWIT algorithm is highly parallelizable, which would enable fast, distributed implementations ideal for big data analysis. Implications to other methods such as expectation maximization PCA (EM PCA) are also presented. Based on our results, general recommendations for efficient application of PCA methods are given according to problem size and available computational resources. MPOWIT and all other methods discussed here are implemented and readily available in the open source GIFT software.

Dataset Information

OCMA: Fast, Memory-Efficient Factorization of Prohibitively Large Relationship Matrices.

Publications

OCMA: Fast, Memory-Efficient Factorization of Prohibitively Large Relationship Matrices.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets