Identifying gene–gene interactions that are highly associated with four quantitative lipid traits across multiple cohorts

Rishika De, Shefali S. Verma, Emily Holzinger, Molly Hall, Amber Burt, David S. Carrell, David R. Crosslin, Gail P. Jarvik, Helena Kuivaniemi, Iftikhar J. Kullo, Leslie A. Lange, Matthew B. Lanktree, Eric B. Larson, Kari E. North, Alex P. Reiner, Vinicius Tragante, Gerard Tromp, James G. Wilson, Folkert W. Asselbergs, Fotios DrenosJason H. Moore, Marylyn D. Ritchie, Brendan Keating, Diane Gilbert-Diamond

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Genetic loci explain only 25–30 % of the heritability observed in plasma lipid traits. Epistasis, or gene–gene interactions may contribute to a portion of this missing heritability. Using the genetic data from five NHLBI cohorts of 24,837 individuals, we combined the use of the quantitative multifactor dimensionality reduction (QMDR) algorithm with two SNP-filtering methods to exhaustively search for SNP–SNP interactions that are associated with HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG). SNPs were filtered either on the strength of their independent effects (main effect filter) or the prior knowledge supporting a given interaction (Biofilter). After the main effect filter, QMDR identified 20 SNP–SNP models associated with HDL-C, 6 associated with LDL-C, 3 associated with TC, and 10 associated with TG (permutation P value <0.05). With the use of Biofilter, we identified 2 SNP–SNP models associated with HDL-C, 3 associated with LDL-C, 1 associated with TC and 8 associated with TG (permutation P value <0.05). In an independent dataset of 7502 individuals from the eMERGE network, we replicated 14 of the interactions identified after main effect filtering: 11 for HDL-C, 1 for LDL-C and 2 for TG. We also replicated 23 of the interactions found to be associated with TG after applying Biofilter. Prior knowledge supports the possible role of these interactions in the genetic etiology of lipid traits. This study also presents a computationally efficient pipeline for analyzing data from large genotyping arrays and detecting SNP–SNP interactions that are not primarily driven by strong main effects.

Original languageEnglish (US)
Pages (from-to)165-178
Number of pages14
JournalHuman genetics
Volume136
Issue number2
DOIs
StatePublished - Feb 1 2017

Fingerprint

Triglycerides
LDL Cholesterol
HDL Cholesterol
Lipids
Multifactor Dimensionality Reduction
Cholesterol
LDL Lipoproteins
Single Nucleotide Polymorphism
National Heart, Lung, and Blood Institute (U.S.)
Genetic Loci

All Science Journal Classification (ASJC) codes

  • Genetics
  • Genetics(clinical)

Cite this

De, Rishika ; Verma, Shefali S. ; Holzinger, Emily ; Hall, Molly ; Burt, Amber ; Carrell, David S. ; Crosslin, David R. ; Jarvik, Gail P. ; Kuivaniemi, Helena ; Kullo, Iftikhar J. ; Lange, Leslie A. ; Lanktree, Matthew B. ; Larson, Eric B. ; North, Kari E. ; Reiner, Alex P. ; Tragante, Vinicius ; Tromp, Gerard ; Wilson, James G. ; Asselbergs, Folkert W. ; Drenos, Fotios ; Moore, Jason H. ; Ritchie, Marylyn D. ; Keating, Brendan ; Gilbert-Diamond, Diane. / Identifying gene–gene interactions that are highly associated with four quantitative lipid traits across multiple cohorts. In: Human genetics. 2017 ; Vol. 136, No. 2. pp. 165-178.
@article{a093fb943a8d4486ad8e58c9cb90cfb8,
title = "Identifying gene–gene interactions that are highly associated with four quantitative lipid traits across multiple cohorts",
abstract = "Genetic loci explain only 25–30 {\%} of the heritability observed in plasma lipid traits. Epistasis, or gene–gene interactions may contribute to a portion of this missing heritability. Using the genetic data from five NHLBI cohorts of 24,837 individuals, we combined the use of the quantitative multifactor dimensionality reduction (QMDR) algorithm with two SNP-filtering methods to exhaustively search for SNP–SNP interactions that are associated with HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG). SNPs were filtered either on the strength of their independent effects (main effect filter) or the prior knowledge supporting a given interaction (Biofilter). After the main effect filter, QMDR identified 20 SNP–SNP models associated with HDL-C, 6 associated with LDL-C, 3 associated with TC, and 10 associated with TG (permutation P value <0.05). With the use of Biofilter, we identified 2 SNP–SNP models associated with HDL-C, 3 associated with LDL-C, 1 associated with TC and 8 associated with TG (permutation P value <0.05). In an independent dataset of 7502 individuals from the eMERGE network, we replicated 14 of the interactions identified after main effect filtering: 11 for HDL-C, 1 for LDL-C and 2 for TG. We also replicated 23 of the interactions found to be associated with TG after applying Biofilter. Prior knowledge supports the possible role of these interactions in the genetic etiology of lipid traits. This study also presents a computationally efficient pipeline for analyzing data from large genotyping arrays and detecting SNP–SNP interactions that are not primarily driven by strong main effects.",
author = "Rishika De and Verma, {Shefali S.} and Emily Holzinger and Molly Hall and Amber Burt and Carrell, {David S.} and Crosslin, {David R.} and Jarvik, {Gail P.} and Helena Kuivaniemi and Kullo, {Iftikhar J.} and Lange, {Leslie A.} and Lanktree, {Matthew B.} and Larson, {Eric B.} and North, {Kari E.} and Reiner, {Alex P.} and Vinicius Tragante and Gerard Tromp and Wilson, {James G.} and Asselbergs, {Folkert W.} and Fotios Drenos and Moore, {Jason H.} and Ritchie, {Marylyn D.} and Brendan Keating and Diane Gilbert-Diamond",
year = "2017",
month = "2",
day = "1",
doi = "10.1007/s00439-016-1738-7",
language = "English (US)",
volume = "136",
pages = "165--178",
journal = "Human Genetics",
issn = "0340-6717",
publisher = "Springer Verlag",
number = "2",

}

De, R, Verma, SS, Holzinger, E, Hall, M, Burt, A, Carrell, DS, Crosslin, DR, Jarvik, GP, Kuivaniemi, H, Kullo, IJ, Lange, LA, Lanktree, MB, Larson, EB, North, KE, Reiner, AP, Tragante, V, Tromp, G, Wilson, JG, Asselbergs, FW, Drenos, F, Moore, JH, Ritchie, MD, Keating, B & Gilbert-Diamond, D 2017, 'Identifying gene–gene interactions that are highly associated with four quantitative lipid traits across multiple cohorts', Human genetics, vol. 136, no. 2, pp. 165-178. https://doi.org/10.1007/s00439-016-1738-7

Identifying gene–gene interactions that are highly associated with four quantitative lipid traits across multiple cohorts. / De, Rishika; Verma, Shefali S.; Holzinger, Emily; Hall, Molly; Burt, Amber; Carrell, David S.; Crosslin, David R.; Jarvik, Gail P.; Kuivaniemi, Helena; Kullo, Iftikhar J.; Lange, Leslie A.; Lanktree, Matthew B.; Larson, Eric B.; North, Kari E.; Reiner, Alex P.; Tragante, Vinicius; Tromp, Gerard; Wilson, James G.; Asselbergs, Folkert W.; Drenos, Fotios; Moore, Jason H.; Ritchie, Marylyn D.; Keating, Brendan; Gilbert-Diamond, Diane.

In: Human genetics, Vol. 136, No. 2, 01.02.2017, p. 165-178.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Identifying gene–gene interactions that are highly associated with four quantitative lipid traits across multiple cohorts

AU - De, Rishika

AU - Verma, Shefali S.

AU - Holzinger, Emily

AU - Hall, Molly

AU - Burt, Amber

AU - Carrell, David S.

AU - Crosslin, David R.

AU - Jarvik, Gail P.

AU - Kuivaniemi, Helena

AU - Kullo, Iftikhar J.

AU - Lange, Leslie A.

AU - Lanktree, Matthew B.

AU - Larson, Eric B.

AU - North, Kari E.

AU - Reiner, Alex P.

AU - Tragante, Vinicius

AU - Tromp, Gerard

AU - Wilson, James G.

AU - Asselbergs, Folkert W.

AU - Drenos, Fotios

AU - Moore, Jason H.

AU - Ritchie, Marylyn D.

AU - Keating, Brendan

AU - Gilbert-Diamond, Diane

PY - 2017/2/1

Y1 - 2017/2/1

N2 - Genetic loci explain only 25–30 % of the heritability observed in plasma lipid traits. Epistasis, or gene–gene interactions may contribute to a portion of this missing heritability. Using the genetic data from five NHLBI cohorts of 24,837 individuals, we combined the use of the quantitative multifactor dimensionality reduction (QMDR) algorithm with two SNP-filtering methods to exhaustively search for SNP–SNP interactions that are associated with HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG). SNPs were filtered either on the strength of their independent effects (main effect filter) or the prior knowledge supporting a given interaction (Biofilter). After the main effect filter, QMDR identified 20 SNP–SNP models associated with HDL-C, 6 associated with LDL-C, 3 associated with TC, and 10 associated with TG (permutation P value <0.05). With the use of Biofilter, we identified 2 SNP–SNP models associated with HDL-C, 3 associated with LDL-C, 1 associated with TC and 8 associated with TG (permutation P value <0.05). In an independent dataset of 7502 individuals from the eMERGE network, we replicated 14 of the interactions identified after main effect filtering: 11 for HDL-C, 1 for LDL-C and 2 for TG. We also replicated 23 of the interactions found to be associated with TG after applying Biofilter. Prior knowledge supports the possible role of these interactions in the genetic etiology of lipid traits. This study also presents a computationally efficient pipeline for analyzing data from large genotyping arrays and detecting SNP–SNP interactions that are not primarily driven by strong main effects.

AB - Genetic loci explain only 25–30 % of the heritability observed in plasma lipid traits. Epistasis, or gene–gene interactions may contribute to a portion of this missing heritability. Using the genetic data from five NHLBI cohorts of 24,837 individuals, we combined the use of the quantitative multifactor dimensionality reduction (QMDR) algorithm with two SNP-filtering methods to exhaustively search for SNP–SNP interactions that are associated with HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG). SNPs were filtered either on the strength of their independent effects (main effect filter) or the prior knowledge supporting a given interaction (Biofilter). After the main effect filter, QMDR identified 20 SNP–SNP models associated with HDL-C, 6 associated with LDL-C, 3 associated with TC, and 10 associated with TG (permutation P value <0.05). With the use of Biofilter, we identified 2 SNP–SNP models associated with HDL-C, 3 associated with LDL-C, 1 associated with TC and 8 associated with TG (permutation P value <0.05). In an independent dataset of 7502 individuals from the eMERGE network, we replicated 14 of the interactions identified after main effect filtering: 11 for HDL-C, 1 for LDL-C and 2 for TG. We also replicated 23 of the interactions found to be associated with TG after applying Biofilter. Prior knowledge supports the possible role of these interactions in the genetic etiology of lipid traits. This study also presents a computationally efficient pipeline for analyzing data from large genotyping arrays and detecting SNP–SNP interactions that are not primarily driven by strong main effects.

UR - http://www.scopus.com/inward/record.url?scp=84995489804&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84995489804&partnerID=8YFLogxK

U2 - 10.1007/s00439-016-1738-7

DO - 10.1007/s00439-016-1738-7

M3 - Article

C2 - 27848076

AN - SCOPUS:84995489804

VL - 136

SP - 165

EP - 178

JO - Human Genetics

JF - Human Genetics

SN - 0340-6717

IS - 2

ER -