University of Groningen
Core gene identification using gene expression
Claringbould, Annique
DOI:
10.33612/diss.145227875
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Claringbould, A. (2020). Core gene identification using gene expression. University of Groningen.
https://doi.org/10.33612/diss.145227875
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
2
The genetic architecture
of molecular traits
Current Opinion in Systems Biology
2017, 1:25–31
doi.org/10.1016/j.coisb.2017.01.002
Annique Claringbould*, Niek de Klein*, Lude Franke
Abstract
Most diseases have both an environmental and genetic component. Although many diseases are strongly heritable, individual genetic variants typically confer only a small effect on disease, and thus these diseases are strongly polygenic. Paradoxically, molecular traits, such as gene expression, methylation, protein or metabolite levels, typically have a lower heritability, but sometimes individual genetic variants show much higher effect sizes on these traits. In this review we discuss the genetic architecture of these molecular traits, and contrast this to the genetic architecture of complex diseases, and provide explanations why strong effects of individual genetic variants on molecular traits do not necessarily need to translate into increased risk of disease.
Graphical abstract The effect size of SNPs decreases when the complexity of a trait increases from very direct consequences, like cis-regulation to more complex phenotypes, like disease susceptibility.
1
2
3
4
5
6
7
8
Introduction
Since most diseases, phenotypes, and molecular traits are heritable, their genetic
architecture is a long-standing question. Are most phenotypes caused by a limited number of large effect variants, or are they due to many variants that each have a small effect? In this review, we compare detected effect size and allele frequencies of associated variants from genome-wide association studies (GWAS) on complex traits and diseases with results from expression- and methylation quantitative trait locus (eQTL and meQTL) studies.
Genetic architecture of common disease: few large effects
Over the last few years, GWAS have shown that associated variants typically only explain a small proportion of the disease variation seen in the majority of common diseases, despite a few examples of common variants with a large effect being found for immune-related traits, particularly in the HLA-region (Daly et al., 2009). Currently, variants with an odds ratio (OR) > 10 are considered ‘highly unusual’ (McCarthy et al., 2008). These observations led to an extensive debate about the genetic architecture of complex diseases (Hansen, 2006; McCarthy et al., 2008; Devlin and Scherer, 2012; Loh et al., 2015; Fuchsberger et al., 2016), including issues like the number, frequency and effect size of associated genetic variants, as well as the degree of shared genetic background with other traits (Mackay, 2001). For complex diseases the architecture is by no means uniform: the number of genes and their effect sizes differ widely (Loh et al., 2015; Fuchsberger et al., 2016; Hou et al., 2016; van Rheenen et al., 2016). However, there is evidence for a shared genetic basis among many diseases (Zhernakova et al., 2009; Pickrell et al., 2016; Shi et al., 2016), and the genetic architecture of most complex disease seems to be highly polygenic.
A much cited interpretation of the overall genetic architecture of diseases places genetic variants in five different groups, based on their allele frequency and effect size or
penetrance (Figure 2.1A, adapted from (McCarthy et al., 2008; Manolio et al., 2009)). In this
representation, complex diseases are characterized by many common genetic variants with small effect sizes, whereas Mendelian diseases are caused by rare variants with large effects. Subsequently, methods were developed that can infer the variance explained by using all directly genotyped SNPs, including those that do not attain genome-wide significance. For complex phenotypes such as height it was established that a considerable proportion of the heritability could be explained by common SNPs, suggesting a highly polygenic genetic architecture (Yang et al., 2010).
Indeed, an inventory of the binary traits in the GWAS Catalog (v1.01, r2016-06-12, p<5×10 -8, Supplementary Note 2.1) reveals that most identified SNPs are common (minor allele frequency (MAF) > 0.1) and have small effect sizes (OR between 1.0 and 1.2, Figure 2.1B).
intermediate frequency variants (0.005 < MAF < 0.1) with larger effects on disease. However, even very large studies have not yet been able to identify many of these hypothesized large effect variants (Fuchsberger et al., 2016) (Figure 2.1B), while recent studies show that they
can be more successfully detected by whole-genome sequencing (Del-Aguila et al., 2015; Walter et al., 2015).
Genetic architecture of molecular traits
Surprisingly, the genetic architecture of complex phenotypic (disease) traits differs substantially from the genetic architecture of molecular traits such as gene expression, methylation, or protein levels. Although the genetic architecture of molecular traits can also be polygenic, a single SNP can often explain a considerable part of the heritability compared to disease phenotypes, whereas the heritability of gene expression, methylation or protein levels is typically lower than complex diseases (Wright et al., 2014; Polderman et al., 2015).
Large effect-sizes of SNPs affecting molecular traits
Genetic variation influences the risk of developing a complex disease through several molecular traits such as gene expression (Emilsson et al., 2008) and methylation (Conerly and Grady, 2010). Investigating eQTLs (Pickrell et al., 2010; Lappalainen et al., 2013; Westra et al., 2013; Gibson, Powell and Marigorta, 2015; Zhernakova et al., 2015) and allele-specific expression (ASE) (Castel et al., 2015; Deelen et al., 2015; Pirinen et al., 2015; Rivas et al., 2015) can characterize the effect of common and rare genetic variants, respectively, on gene expression. Similarly, intermediate frequency and common SNPs that affect methylation levels at CpG sites can be detected by meQTL mapping (Bonder et al., 2015).
Analogous to the genetic architecture of common diseases presented in Figure 2.1B, the
genetic architecture of gene expression and methylation may be represented by plotting the allele frequency of QTLs against their effect size (Figure 2.1C-D, Supplementary Note 2.2).
We first compared the proportion of variance explained for cis-eQTLs identified in RNA-seq data in blood (BIOS consortium, N=2,116, only SNPs tested with MAF > 0.05) (Zhernakova et al., 2015) and in Epstein-Barr virus-transformed lymphoblastoid cell lines (GEUVADIS consortium, N=373, only SNPs tested with MAF > 0.05) (Lappalainen et al., 2013). Comparing the two cohorts indicates that sample size has an impact on the number of identified eQTL SNPs, but not on their effect size (Figure 2.1C).
1
2
3
4
5
6
7
8
Figure 2.1 (A) Proposed genetic architecture of diseases, adapted from (McCarthy et al., 2008; Manolio et al., 2009). (B) Minor allele frequency (MAF) set out against odds ratio (OR) of genome-wide significant GWAS SNPs. The data is downloaded from the GWAS catalog (Supplementary Note 2.1). Histograms on the right and at the top indicate the frequency distribution of the SNPs, the dot size indicates the total sample size of the GWAS, and the color represents the year of publication. Rare and intermediate frequency variants (MAF < 0.1) have a higher OR on average. (C) MAF against variance explained (R2) of cis expression quantitative trait locus (eQTL) SNPs. Light blue SNPs are from the GEUVADIS
consortium (N=373), dark blue SNPs from the BIOS consortium (N=2116) (Supplementary Note 2.2). The plots on the top and right of the Figure illustrate the density distribution of SNPs from both cohorts. Despite different sample sizes, the distribution of the cis eQTL SNPs is similar for both cohorts. (D) MAF against variance explained (R2) of cis and trans
methylation quantitative trait locus (eQTL) SNPs. Dark blue SNPs are cis meQTLs and light blue SNPs are trans meQTLs (Supplementary Note 2.3). Histograms on the right and at the top indicate the frequency distribution of the SNPs. Common SNPs often have large effect sizes, and trans effects explain much less methylation variation on average.
Notably, the effects of many eQTLs and meQTLs are large, especially compared to the degree of heritability of a disease explained by GWAS: some eQTLs explain as much as 80% of the variation at transcript level (Figure 2.1C) and many cis-meQTLs explain over 70% of the
methylation level variation (Figure 2.1D). While common GWAS SNPs only rarely have a large
effect on the phenotype, the impact of common SNPs on molecular traits can thus be much greater.
Another observation is that the minor allele frequency distribution of disease variants (Figure 2.1B) is different from the MAF distribution of both eQTL and meQTLs (Figure 2.1C-D): molecular traits are influenced by variants with on average a lower MAF, as compared to
complex traits for which the average MAF is higher.
Many independent SNPs affecting the same molecular trait
While for many complex diseases many independent associations have so-far been found, the number of independent associations for molecular traits has been studied less well. Conditional analyses (i.e. correction for primary cis-eQTLs) can be performed to ascertain whether secondary signals can be identified. Zhernakova et al. (2015) recently performed such an analysis and observed that more than half of the transcripts are governed by more than one cis-eQTL effect, suggesting that the genetic architecture of gene expression regulation for most genes is polygenic, similar to complex diseases.
Local and distal effects
However, these conditional analyses have so-far only identified multiple independent SNPs that are working in cis: SNPs most often affect nearby molecular traits in cis through mechanisms such as transcription factor binding disruption in regulatory regions or promotor disruption within transcription start sites (Brown, Mangravite and Engelhardt, 2013). To identify genetic variants that are distantly located, trans-QTL mapping can be conducted. Those trans effects are hypothesized to be mediated by multiple cis effects and complex downstream regulation on a molecular trait (Westra et al., 2013; Wong et al., 2016), and are identified less frequently (Figure 2.2), due to severe multiple-testing issues when
comparing millions of genetic variants with tens of thousands of molecular traits. So-far the largest QTL studies have observed a trans-acting proportion of 4.5% (Westra et al., 2013) and 2.4% (Wright et al., 2014) of total eQTLs, and 3.6% (Bonder et al., 2015) and 6.5% for meQTLs (Gaunt et al., 2016). With increasing sample-sizes these estimates will likely go up in the near future, since it has been estimated that cis-effects explain only 23% of the total heritability of gene expression level regulation (Wright et al., 2014), thus suggesting that distal effects strongly outnumber local effects. However, for both eQTLs and meQTLs, cis effect sizes are on average several times higher than trans effects (Figures 2.1D, 2.3A), which is likely due to
1
2
3
4
5
6
7
8
With the current sample-sizes, the percentage of tested genes with a significant QTL is higher for cis than for trans (Figure 2.3A). This will undoubtedly change when sample-sizes increase
further. However, what can be concluded now is that the proportion of CpG sites that show trans-meQTL effects is lower than the proportion of genes with a trans-eQTL effect. When accounting for the different numbers of samples in trans-eQTL and trans-meQTL studies this pictue does not change. This indicates that gene expression levels are more polygenic than methylation levels.
Figure 2.2 Pie charts comparing the number of observed eQTLs (A) and meQTLs (B) based on (Westra et al., 2013; Bonder et al., 2015). Blue indicates cis effects, red are the trans effects.
Cell-type and context-specificity of QTLs
It should be noted, however, that the effects of genetic variants on molecular traits are highly variable: it is now clear that the effects of genetic variation on gene expression levels can depend strongly on cell type, tissue type, and context, such as stimulations by pathogens (Fu et al., 2012; Gat-Viks et al., 2013; Fairfax et al., 2014; Lee et al., 2014; Melé et al., 2015). Indeed, blood cis-eQTLs are often inconsistent in other tissues: blood QTLs may be absent in a different tissue, they may have a different effect size, or even show an opposite allelic effect (Fu et al., 2012). This is even more pronounced for trans-eQTLs: some trans-eQTLs can only be replicated in a specific cell type (Westra et al., 2013). Tissue-dependent eQTL effects are enriched in SNPs associated with complex traits (Fu et al., 2012), while the cell-type specific
effects identified in the HLA region (Fairfax et al., 2012) are just one example of how tissue specificity can influence the risk of developing the disease. Surprisingly, trans-meQTLs are much more stable across cell types (Gutierrez-Arcelus et al., 2015): it has been shown that the far majority of whole blood trans -meQTLs replicate in lymphocytes (Bonder et al., 2015), suggesting that meQTLs are generally very stable. It should be noted though, exceptions exist: there are tissue-specific methylation patterns (Lokk et al., 2014) and cis-meQTLs have been found whose CpG levels show correlations with expression levels of tissue-specific alternatively spliced exons (Gutierrez-Arcelus et al., 2015).
Genetic architecture of other protein and metabolite levels
Eventually changes in methylation and particularly gene expression levels due to genetic variation should show downstream consequences. Since gene expression levels are not a direct proxy for protein levels (Parts et al., 2014; Liu and Aebersold, 2016), protein quantitative trait loci (pQTLs) can give a more accurate measurement of the effect of SNPs on protein abundance. Interestingly, not all pQTLs overlap with eQTLs, suggesting that multiple mechanisms regulate protein levels (Wu et al., 2013; Liu et al., 2015). Recent research on mice, measuring genome-wide gene expression and protein levels on the same samples, suggests that distal pQTLs act on protein levels through mediator proteins and post-transcriptional mechanisms, while local pQTL SNPs directly influence the protein level through gene expression regulation (Chick et al., 2016).
Cytokines are a specific class of proteins: they play an important role in immunological disorders. Cytokine abundances are especially variable in response to pathogens, and they can be mapped to SNPs to find stimulation-induced cytokine quantitative trait loci (cQTLs) (Lu et al., 2011; Li et al., 2016). While some cQTLs are cytokine-specific, other QTLs are shared among cytokines (Li et al., 2016).
Likewise, it is possible to detect SNPs that influence nuclear magnetic resonance (NMR)-derived metabolite abundances in urine or plasma (Nicholson et al., 2011; Shin et al., 2014). These metabolite quantitative trait loci (mQTLs) show that metabolite levels, like protein levels, are highly heritable, with some variants that have been detected explaining more than 40% of metabolite level variation (Nicholson et al., 2011).
As sample sizes of the metabolite and protein QTL studies are still small, they are difficult to compare to the large-scale eQTL and meQTL studies that have been conducted so-far. Human pQTL studies have detected around 180 pQTLs with a large impact (Wu et al., 2013): these signals are the ones most easily detected. However, a recent study in mice pQTLs (Chick et al., 2016) showed that fewer pQTLs than eQTLs were detected in liver when using the same mice for both analyses (N=192, FDR < 0.01, Figure 2.3B). Although mice
1
2
3
4
5
6
7
8
and humans have different physiologies, the finding suggests that human pQTLs are more difficult to detect due to their smaller average effect size. In the future, human pQTL studies with a larger number of samples as compared to eQTL and meQTL studies are therefore needed in order to identify protein and metabolite QTLs that have smaller effect sizes.
Figure 2.3 (A) Percentage of total measured SNPs detected as cis eQTL, cis meQTL, trans eQTL and trans meQTL SNPs, based on (Westra et al., 2013; Bonder et al., 2015; Zhernakova et al., 2015). Cis effects are identified for a much larger percentage of SNPs compared to trans effects. (B) Percentage of total measured SNPs detected as cis eQTL, and cis protein QTL (pQTL) SNPs in mice, based on (Chick et al., 2016). For the same mice and the same number of transcripts and proteins, eQTLs are more often detected than pQTLs.
Conclusion
In contrast to GWAS SNPs, SNPs affecting molecular traits can have a very large effect size. Paradoxically, the twin estimated heritability of expression levels (h2=0.142) (Wright et al.,
2014) is lower than the twin estimated heritability of diseases (h2=0.593) (Polderman et al.,
2015). Given the strong effect sizes that are often observed for molecular traits, in contrast to complex traits, this suggests that the genetic architecture of molecular traits is much less polygenic than that of complex diseases.
However, it remains elusive how SNPs can have such a strong effect on a single molecular trait, while they generally have only a small effect on disease phenotypes. One possibility is that the affected molecular trait does not play such an important role. This would explain
why there are common SNPs with high effect sizes. Another scenario is that molecular trait levels are redundant: variability of one trait (e.g. the expression levels of a specific gene) may only become important in the absence of the other trait (e.g. another gene that has the same biological function). Alternatively, changes in levels of a single molecular trait need not have a big effect downstream because many proteins are involved in the regulation of a single gene. The abundance of transcripts regulated by more than one cis-eQTL in combination with the considerable number of trans-acting effects supports the hypothesis that complex pathways buffer the large variation in molecular traits to result in a small disease-inducing effect.
To our knowledge, except for some allele-specific analyses, few QTL studies have so far looked at the molecular effects of rare variants (MAF < 0.005), which makes it hard to predict what effects these variants will have on molecular traits. However, considering that the intermediate variants show an increase in their effect size compared to common variants, and rare SNPs are under more evolutionary pressure, we would expect some of these rare variants to have a very strong effect on molecular traits.
To better understand disease mechanisms, we need to gain a more complete picture of the genetic architecture of molecular traits. If rare variants do show an even larger effect on molecular traits than intermediate or common variants, it is key that QTL studies should investigate them using larger sample sizes, as well as using tools designed to identify the effects from rare SNPs, such as allele-specific analysis.
Acknowledgements
We thank Jackie Senior for editing the final text. This work is supported by a grant from the European Research Counsil (ERC Starting Grant agreement number 637640 ImmRisk) to Lude Franke and an NWO-VIDI grant 917-14374 from the Netherlands Organization for Scientific Research.
License and version
This article was previously published in Current Opinion in Systems Biology (doi. org/10.1016/j.coisb.2017.01.002). No changes were made to the original publication.
1
2
3
4
5
6
7
8
Supplementary Material
Supplementary note 2.1The GWAS catalog (v1.01, r2016-06-12) data was downloaded on 01-07-2016 from https:// www.ebi.ac.uk/gwas/docs/downloads. We determined which studies reported ρ-values by checking if there was a unit of measurement in the ‘OR or BETA’ column. These studies were excluded to keep the binary trait studies. The data was subsequently filtered to include only genome-wide significant SNPs with a p-value < 5×10-8, and with a reported confidence interval (CI) around the odds ratio (OR), single value for risk allele frequency, and sample size.
Supplementary note 2.2
The BIOS consortium eQTL data (v.19-11-2015) was downloaded on 01-07-2016 from http:// genenetwork.nl/biosqtlbrowser/. The GEUVADIS consortium eQTL data were retrieved from the replication in Zhernakova et al., 2015, bioRxiv. We added the allele frequency information based on GoNL whole-genome sequence data (release 5, https://molgenis26.target.rug. nl/downloads/gonl_public/variants/release5/) to both sets, and confined the plots to SNPs with a minor allele frequency of > 0.05, as used in the eQTL analyses of the BIOS and GEUVADIS consortia.
Supplementary note 2.3
The cis and trans meQTL data (v.1-9-2015) was downloaded on 07-07-2016 from http:// genenetwork.nl/biosqtlbrowser/. We added the allele frequency information based on GoNL whole-genome sequence data (release 5, https://molgenis26.target.rug.nl/ downloads/gonl_public/variants/release5/) to both sets, and confined the plots to SNPs with a minor allele frequency of > 0.05, as used in the meQTL analyses.
References
Albert, F. W. and Kruglyak, L. (2015) ‘The role of regulatory variation in complex traits and disease’, Nature Reviews Genetics. Nature Publishing Group, 16(4), pp. 197–212. doi: 10.1038/nrg3891.
Bonder, M. J. et al. (2015) ‘Disease variants alter transcription factor levels and methylation of their binding sites’, bioRxiv. doi: 10.1101/033084.
Brown, C. D., Mangravite, L. M. and Engelhardt, B. E. (2013) ‘Integrative modeling of eQTLs and cis-regulatory elements suggests mechanisms underlying cell type specificity of eQTLs.’, PLoS genetics, 9(8), p. e1003649. doi: 10.1371/journal.pgen.1003649.
Castel, S. E. et al. (2015) ‘Tools and best practices for allelic expression analysis’, bioRxiv. Genome Biology, p. 016097. doi: 10.1101/016097.
Chick, J. M. et al. (2016) ‘Defining the consequences of genetic variation on a proteome-wide scale’, Nature, 534, pp. 500–505. doi: 10.1038/nature18270.
Conerly, M. and Grady, W. M. (2010) ‘Insights into the role of DNA methylation in disease through the use of mouse models.’, Disease models & mechanisms, 3(5–6), pp. 290–297. doi: 10.1242/dmm.004812.
Daly, A. K. et al. (2009) ‘HLA-B*5701 genotype is a major determinant of drug-induced liver injury due to flucloxacillin.’, Nature genetics, 41(7), pp. 816–819. doi: 10.1038/ng.379.
Deelen, P. et al. (2015) ‘Calling genotypes from public RNA-sequencing data enables identification of genetic variants that affect gene-expression levels’, Genome Medicine, 7(1), p. 30. doi: 10.1186/s13073-015-0152-4. Del-Aguila, J. L. et al. (2015) ‘Alzheimer’s disease: rare variants with large effect sizes’, Current Opinion in Genetics & Development, 33, pp. 49–55. doi: 10.1016/j.gde.2015.07.008.
Devlin, B. and Scherer, S. W. (2012) ‘Genetic architecture in autism spectrum disorder.’, Current opinion in genetics & development, 22(3), pp. 229–37. doi: 10.1016/j.gde.2012.03.002.
Emilsson, V. et al. (2008) ‘Genetics of gene expression and its effect on disease’, Nature. Nature Publishing Group, 452(7186), pp. 423–428. doi: 10.1038/nature06758.
Fairfax, B. P. et al. (2012) ‘Genetics of gene expression in primary immune cells identifies cell type–specific master regulators and roles of HLA alleles’, Nature Genetics, 44(5), pp. 502–510. doi: 10.1038/ng.2205.
Fairfax, B. P. et al. (2014) ‘Innate immune activity conditions the effect of regulatory variants upon monocyte gene expression.’, Science (New York, N.Y.). American Association for the Advancement of Science, 343(6175), p. 1246949. doi: 10.1126/science.1246949.
Fu, J. et al. (2012) ‘Unraveling the regulatory mechanisms underlying tissue-dependent genetic variation of gene expression’, PLoS Genetics. Edited by G. Gibson. Public Library of Science, 8(1), p. e1002431. doi: 10.1371/journal. pgen.1002431.
Fuchsberger, C. et al. (2016) ‘The genetic architecture of type 2 diabetes’, Nature. Nature Publishing Group, 536, pp. 41–47. doi: 10.1038/nature18642.
Gat-Viks, I. et al. (2013) ‘Deciphering molecular circuits from genetic variation underlying transcriptional responsiveness to stimuli’, Nature Biotechnology. Nature Research, 31(4), pp. 342–349. doi: 10.1038/nbt.2519. Deciphering.
Gaunt, T. R. et al. (2016) ‘Systematic identification of genetic influences on methylation across the human life course’, Genome Biology. Genome Biology, 17(1), p. 61. doi: 10.1186/s13059-016-0926-z.
Gibson, G., Powell, J. E. and Marigorta, U. M. (2015) ‘Expression quantitative trait locus analysis for translational medicine’, Genome Medicine. Genome Medicine, 7, p. 60. doi: 10.1186/s13073-015-0186-7.
Gutierrez-Arcelus, M. et al. (2015) ‘Tissue-specific effects of genetic and epigenetic variation on gene regulation and splicing’, PLoS genetics. Edited by C. D. Brown. Public Library of Science, 11(1), p. e1004958. doi: 10.1371/ journal.pgen.1004958.
Hansen, T. E. (2006) ‘The Evolution of Genetic Architecture’, Annu. Rev. Ecol. Evol. Syst., 37, pp. 123–157. doi: 10.2307/annurev.ecolsys.37.091305.30000007.
Hou, J. et al. (2016) ‘The hidden complexity of Mendelian traits across natural yeast populations’, Cell Reports. Elsevier, 16, pp. 1106–1114. doi: 10.1016/j.celrep.2016.06.048.
Lappalainen, T. et al. (2013) ‘Transcriptome and genome sequencing uncovers functional variation in humans’, Nature. Nature Publishing Group, 501(7468), pp. 506–511. doi: 10.1038/nature12531.
Lee, M. N. et al. (2014) ‘Common genetic variants modulate pathogen-sensing responses in human dendritic cells’, Science (New York, N.Y.). NIH Public Access, 343(6175), p. 1246980. doi: 10.1126/science.1246980. Li, Y. et al. (2016) ‘Inter-individual variability and genetic influences on cytokine responses to bacteria and fungi’, Nature Medicine. Nature Research, 22, pp. 952–960. doi: 10.1038/nm.4139.
1
2
3
4
5
6
7
8
Liu, Y. et al. (2015) ‘Quantitative variability of 342 plasma proteins in a human twin population.’, Molecular systems biology. EMBO Press, 11(1), p. 786. doi: 10.15252/msb.20145728.
Liu, Y. and Aebersold, R. (2016) ‘The interdependence of transcript and protein abundance : new data – new complexities’, Molecular systems biology, 12, p. 856.
Loh, P.-R. et al. (2015) ‘Contrasting genetic architectures of schizophrenia and other complex diseases using fast variance-components analysis.’, Nature genetics, 47, pp. 1385–1392. doi: 10.1038/ng.3431.
Lokk, K. et al. (2014) ‘DNA methylome profiling of human tissues identifies global and tissue-specific methylation patterns’, Genome biology. BioMed Central, 15(4), p. r54. doi: 10.1186/gb-2014-15-4-r54.
Lu, X. et al. (2011) ‘Mapping quantitative trait loci for cytokines in the pig’, Animal Genetics. Blackwell Publishing Ltd, 42(1), pp. 1–5. doi: 10.1111/j.1365-2052.2010.02071.x.
Mackay, T. (2001) ‘The Genetic Architecture of Quantitative Traits’, Annual Reviews Genetics, 35, pp. 303–339. Manolio, T. A. et al. (2009) ‘Finding the missing heritability of complex diseases.’, Nature, 461(7265), pp. 747–53. doi: 10.1038/nature08494.
McCarthy, M. I. et al. (2008) ‘Genome-wide association studies for complex traits: consensus, uncertainty and challenges.’, Nature reviews. Genetics, 9(5), pp. 356–369. doi: 10.1038/nrg2344.
Melé, M. et al. (2015) ‘The human transcriptome across tissues and individuals.’, Science (New York, N.Y.). American Association for the Advancement of Science, 348(6235), pp. 660–665. doi: 10.1126/science.aaa0355. Nicholson, G. et al. (2011) ‘A genome-wide metabolic QTL analysis in europeans implicates two Loci shaped by recent positive selection’, PLoS Genetics. Edited by G. S. Barsh. Public Library of Science, 7(9), p. e1002270. doi: 10.1371/journal.pgen.1002270.
Parts, L. et al. (2014) ‘Heritability and genetic basis of protein level variation in an outbred population’, Genome Research, 24(8), pp. 1363–1370. doi: 10.1101/gr.170506.113.
Pickrell, J. K. et al. (2010) ‘Understanding mechanisms underlying human gene expression variation with RNA sequencing’, Nature, 464(7289), pp. 768–772. doi: 10.1038/nature08872.
Pickrell, J. K. et al. (2016) ‘Detection and interpretation of shared genetic influences on 42 human traits’, Nature genetics, 48, pp. 709–717. doi: 10.1038/ng.3570.
Pirinen, M. et al. (2015) ‘Assessing allele-specific expression across multiple tissues from RNA-seq read data.’, Bioinformatics. Oxford University Press, 31, pp. 2497–2504. doi: 10.1093/bioinformatics/btv074.
Polderman, T. J. C. et al. (2015) ‘Meta-analysis of the heritability of human traits based on fifty years of twin studies’, Nature Genetics. Nature Publishing Group, 47(7), pp. 702–709. doi: 10.1038/ng.3285.
van Rheenen, W. et al. (2016) ‘Genome-wide association analyses identify new risk variants and the genetic architecture of amyotrophic lateral sclerosis’, Nature Genetics. Nature Research. doi: 10.1038/ng.3622. Rivas, M. A. et al. (2015) ‘Effect of predicted protein-truncating genetic variants on the human transcriptome’, Science (New York, N.Y.). American Association for the Advancement of Science, 348, pp. 666–669. doi: 10.1126/ science.1261877.
Shi, H. et al. (2016) ‘Contrasting the genetic architecture of 30 complex traits from summary association data’, The American Journal of Human Genetics, 99(1), pp. 139–153. doi: 10.1016/j.ajhg.2016.05.013.
Shin, S.-Y. et al. (2014) ‘An atlas of genetic influences on human blood metabolites.’, Nature genetics, 46(6), pp. 543–550. doi: 10.1038/ng.2982.
Walter, K. et al. (2015) ‘The UK10K project identifies rare variants in health and disease’, Nature, 526(7571), pp. 82–90. doi: 10.1038/nature14962.
Westra, H.-J. et al. (2013) ‘Systematic identification of trans eQTLs as putative drivers of known disease associations’, Nature Genetics. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved., 45(10), pp. 1238–1243.
Wong, E. S. et al. (2016) ‘Interplay of cis and trans mechanisms driving transcription factor binding, chromatin, and gene expression evolution’, bioRxiv. Cold Spring Harbor Labs Journals. doi: 10.1101/059873.
Wright, F. A. et al. (2014) ‘Heritability and genomics of gene expression in peripheral blood.’, Nature genetics. Nature Publishing Group, 46, pp. 430–437. doi: 10.1038/ng.2951.
Wu, L. et al. (2013) ‘Variation and genetic control of protein abundance in humans’, Nature. Nature Research, 499(7456), pp. 79–82. doi: 10.1038/nature12223.
Yang, J. et al. (2010) ‘Common SNPs explain a large proportion of the heritability for human height’, Nature Genetics. Nature Publishing Group, 42(7), pp. 565–569. doi: 10.1038/ng.608.
Zhernakova, A. et al. (2009) ‘Detecting shared pathogenesis from the shared genetics of immune-related diseases’, Nat. Rev. Genet., 10(1), pp. 43–55. doi: 10.1038/nrg2489.
Zhernakova, D. et al. (2015) ‘Hypothesis-free identification of modulators of genetic risk factors’, bioRxiv. Cold Spring Harbor Labs Journals, p. 033217. doi: 10.1101/033217.
