A novel SCCA approach via truncated â.," 1-norm and truncated group lasso for brain imaging genetics

Lei Du, Kefei Liu, Tuo Zhang, Xiaohui Yao, Jingwen Yan, Shannon L. Risacher, Junwei Han, Lei Guo, Andrew Saykin, Li Shen

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

Motivation Brain imaging genetics, which studies the linkage between genetic variations and structural or functional measures of the human brain, has become increasingly important in recent years. Discovering the bi-multivariate relationship between genetic markers such as single-nucleotide polymorphisms (SNPs) and neuroimaging quantitative traits (QTs) is one major task in imaging genetics. Sparse Canonical Correlation Analysis (SCCA) has been a popular technique in this area for its powerful capability in identifying bi-multivariate relationships coupled with feature selection. The existing SCCA methods impose either the â.," 1-norm or its variants to induce sparsity. The â.," 0-norm penalty is a perfect sparsity-inducing tool which, however, is an NP-hard problem. Results In this paper, we propose the truncated â.," 1-norm penalized SCCA to improve the performance and effectiveness of the â.," 1-norm based SCCA methods. Besides, we propose an efficient optimization algorithms to solve this novel SCCA problem. The proposed method is an adaptive shrinkage method via tuning I.,. It can avoid the time intensive parameter tuning if given a reasonable small I.,. Furthermore, we extend it to the truncated group-lasso (TGL), and propose TGL-SCCA model to improve the group-lasso-based SCCA methods. The experimental results, compared with four benchmark methods, show that our SCCA methods identify better or similar correlation coefficients, and better canonical loading profiles than the competing methods. This demonstrates the effectiveness and efficiency of our methods in discovering interesting imaging genetic associations.

Original languageEnglish (US)
Pages (from-to)278-285
Number of pages8
JournalBioinformatics
Volume34
Issue number2
DOIs
StatePublished - Jan 15 2018

Fingerprint

Canonical Correlation Analysis
Lasso
Neuroimaging
Brain
Imaging
Imaging techniques
Norm
Tuning
Nucleotides
Polymorphism
Feature extraction
Computational complexity
Sparsity
Genetic Association
alachlor
Genetics
Benchmarking
Genetic Variation
Single nucleotide Polymorphism
Parameter Tuning

ASJC Scopus subject areas

  • Statistics and Probability
  • Biochemistry
  • Molecular Biology
  • Computer Science Applications
  • Computational Theory and Mathematics
  • Computational Mathematics

Cite this

A novel SCCA approach via truncated â.," 1-norm and truncated group lasso for brain imaging genetics. / Du, Lei; Liu, Kefei; Zhang, Tuo; Yao, Xiaohui; Yan, Jingwen; Risacher, Shannon L.; Han, Junwei; Guo, Lei; Saykin, Andrew; Shen, Li.

In: Bioinformatics, Vol. 34, No. 2, 15.01.2018, p. 278-285.

Research output: Contribution to journalArticle

Du, Lei ; Liu, Kefei ; Zhang, Tuo ; Yao, Xiaohui ; Yan, Jingwen ; Risacher, Shannon L. ; Han, Junwei ; Guo, Lei ; Saykin, Andrew ; Shen, Li. / A novel SCCA approach via truncated â.," 1-norm and truncated group lasso for brain imaging genetics. In: Bioinformatics. 2018 ; Vol. 34, No. 2. pp. 278-285.
@article{2bd0ca9e2a384317a471a938bb267c9b,
title = "A novel SCCA approach via truncated {\^a}.,{"} 1-norm and truncated group lasso for brain imaging genetics",
abstract = "Motivation Brain imaging genetics, which studies the linkage between genetic variations and structural or functional measures of the human brain, has become increasingly important in recent years. Discovering the bi-multivariate relationship between genetic markers such as single-nucleotide polymorphisms (SNPs) and neuroimaging quantitative traits (QTs) is one major task in imaging genetics. Sparse Canonical Correlation Analysis (SCCA) has been a popular technique in this area for its powerful capability in identifying bi-multivariate relationships coupled with feature selection. The existing SCCA methods impose either the {\^a}.,{"} 1-norm or its variants to induce sparsity. The {\^a}.,{"} 0-norm penalty is a perfect sparsity-inducing tool which, however, is an NP-hard problem. Results In this paper, we propose the truncated {\^a}.,{"} 1-norm penalized SCCA to improve the performance and effectiveness of the {\^a}.,{"} 1-norm based SCCA methods. Besides, we propose an efficient optimization algorithms to solve this novel SCCA problem. The proposed method is an adaptive shrinkage method via tuning I.,. It can avoid the time intensive parameter tuning if given a reasonable small I.,. Furthermore, we extend it to the truncated group-lasso (TGL), and propose TGL-SCCA model to improve the group-lasso-based SCCA methods. The experimental results, compared with four benchmark methods, show that our SCCA methods identify better or similar correlation coefficients, and better canonical loading profiles than the competing methods. This demonstrates the effectiveness and efficiency of our methods in discovering interesting imaging genetic associations.",
author = "Lei Du and Kefei Liu and Tuo Zhang and Xiaohui Yao and Jingwen Yan and Risacher, {Shannon L.} and Junwei Han and Lei Guo and Andrew Saykin and Li Shen",
year = "2018",
month = "1",
day = "15",
doi = "10.1093/bioinformatics/btx594",
language = "English (US)",
volume = "34",
pages = "278--285",
journal = "Bioinformatics",
issn = "1367-4803",
publisher = "Oxford University Press",
number = "2",

}

TY - JOUR

T1 - A novel SCCA approach via truncated â.," 1-norm and truncated group lasso for brain imaging genetics

AU - Du, Lei

AU - Liu, Kefei

AU - Zhang, Tuo

AU - Yao, Xiaohui

AU - Yan, Jingwen

AU - Risacher, Shannon L.

AU - Han, Junwei

AU - Guo, Lei

AU - Saykin, Andrew

AU - Shen, Li

PY - 2018/1/15

Y1 - 2018/1/15

N2 - Motivation Brain imaging genetics, which studies the linkage between genetic variations and structural or functional measures of the human brain, has become increasingly important in recent years. Discovering the bi-multivariate relationship between genetic markers such as single-nucleotide polymorphisms (SNPs) and neuroimaging quantitative traits (QTs) is one major task in imaging genetics. Sparse Canonical Correlation Analysis (SCCA) has been a popular technique in this area for its powerful capability in identifying bi-multivariate relationships coupled with feature selection. The existing SCCA methods impose either the â.," 1-norm or its variants to induce sparsity. The â.," 0-norm penalty is a perfect sparsity-inducing tool which, however, is an NP-hard problem. Results In this paper, we propose the truncated â.," 1-norm penalized SCCA to improve the performance and effectiveness of the â.," 1-norm based SCCA methods. Besides, we propose an efficient optimization algorithms to solve this novel SCCA problem. The proposed method is an adaptive shrinkage method via tuning I.,. It can avoid the time intensive parameter tuning if given a reasonable small I.,. Furthermore, we extend it to the truncated group-lasso (TGL), and propose TGL-SCCA model to improve the group-lasso-based SCCA methods. The experimental results, compared with four benchmark methods, show that our SCCA methods identify better or similar correlation coefficients, and better canonical loading profiles than the competing methods. This demonstrates the effectiveness and efficiency of our methods in discovering interesting imaging genetic associations.

AB - Motivation Brain imaging genetics, which studies the linkage between genetic variations and structural or functional measures of the human brain, has become increasingly important in recent years. Discovering the bi-multivariate relationship between genetic markers such as single-nucleotide polymorphisms (SNPs) and neuroimaging quantitative traits (QTs) is one major task in imaging genetics. Sparse Canonical Correlation Analysis (SCCA) has been a popular technique in this area for its powerful capability in identifying bi-multivariate relationships coupled with feature selection. The existing SCCA methods impose either the â.," 1-norm or its variants to induce sparsity. The â.," 0-norm penalty is a perfect sparsity-inducing tool which, however, is an NP-hard problem. Results In this paper, we propose the truncated â.," 1-norm penalized SCCA to improve the performance and effectiveness of the â.," 1-norm based SCCA methods. Besides, we propose an efficient optimization algorithms to solve this novel SCCA problem. The proposed method is an adaptive shrinkage method via tuning I.,. It can avoid the time intensive parameter tuning if given a reasonable small I.,. Furthermore, we extend it to the truncated group-lasso (TGL), and propose TGL-SCCA model to improve the group-lasso-based SCCA methods. The experimental results, compared with four benchmark methods, show that our SCCA methods identify better or similar correlation coefficients, and better canonical loading profiles than the competing methods. This demonstrates the effectiveness and efficiency of our methods in discovering interesting imaging genetic associations.

UR - http://www.scopus.com/inward/record.url?scp=85040572473&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040572473&partnerID=8YFLogxK

U2 - 10.1093/bioinformatics/btx594

DO - 10.1093/bioinformatics/btx594

M3 - Article

AN - SCOPUS:85040572473

VL - 34

SP - 278

EP - 285

JO - Bioinformatics

JF - Bioinformatics

SN - 1367-4803

IS - 2

ER -