Refining the accuracy of validated target identification through coding variant fine-mapping in type 2 diabetes
Anubha Mahajan, Jennifer Wessel, Sara M. Willems, Zhao Wei, Neil R. Robertson, Audrey Y. Chu, Wei Gan, Hidetoshi Kitajima, Daniel Taliun, N. William Rayner, Xiuqing Guo, Yingchang Lu, Man Li, Richard A. Jensen, Yao Hu, Shaofeng Huo, Kurt K. Lohman, Weihua Zhang, James P. Cook, Bram P. Prins, Jason Flannick, Niels Grarup, Vassily Trubetskoy, Jasmina Kravić, Young Jin Kim, Denis Rybin, Hanieh Yaghootkar, Martina Müller‐Nurasyid, Karina Meidtner, Ruifang Li‐Gao, Tibor V. Varga, Jonathan Marten, Li Jin, Albert V. Smith, Ping An, Symen Ligthart, Stefan Gustafsson, Giovanni Malerba, Ayşe Demirkan, Juan Fernández Tajes, Valgerður Steinthórsdóttir, Matthias Wuttke, Cécile Lecœur, Michael Preuß, Lawrence F. Bielak, Marielisa Graff, Heather M. Highland, Anne E. Justice, Dajiang J. Liu, Eirini Marouli, Gina M. Peloso, Helen R. Warren, Saima Afaq, Shoaib Afzal, Emma Ahlqvist, Peter Almgren, Najaf Amin, Lia B. Bang, Alain G. Bertoni, Cristina Bombieri, Jette Bork‐Jensen, Ivan Brandslund, Jennifer A. Brody, Noël P. Burtt, Mickaël Canouil, Yii‐Der Ida Chen, Yoon Shin Cho, Cramer Christensen, Sophie V. Eastwood, Kai‐Uwe Eckardt, Krista Fischer, Giovanni Gambaro, Vilmantas Giedraitis, Megan L. Grove, Hugoline G. de Haan, Sophie Hackinger, Yang Hai, Sohee Han, Anne Tybjærg‐Hansen, Marie‐France Hivert, Bo Isomaa, Susanne Jäger, Marit E. Jørgensen, Torben Jørgensen, Annemari Käräjämäki, Bong-Jo Kim, Sung‐Soo Kim, Heikki A. Koistinen, Péter Kovács, Jennifer Kriebel, Florian Kronenberg, Kristi Läll, Leslie A. Lange, Jung‐Jin Lee, Benjamin Lehne, Huaixing Li, Keng‐Hung Lin, Allan Linneberg, Ching‐Ti Liu, Jun LiubioRxiv2017 · DOI
This study refines our understanding of genetic risk for Type 2 Diabetes (T2D) by precisely pinpointing causal coding variants. By fine-mapping association signals, the researchers identified 16 T2D-associated coding variants that are likely causal, distinguishing them from "false leads" and paving the way for more accurate target identification for drug development.
What Problem Does This Paper Address?
Identifying the specific genetic variants that cause diseases like Type 2 Diabetes (T2D) is crucial for understanding disease mechanisms and developing new treatments. While Genome-Wide Association Studies (GWAS) have identified thousands of genetic associations with complex traits, most of these signals point to non-coding regions of DNA. Pinpointing the exact causal variant and the gene it affects in these regions is challenging. Coding variants, which directly alter protein sequences, offer a more direct path to biological insight. However, even when a coding variant is associated with a disease, it's not always clear if it's the causal variant or just "hitchhiking" on a nearby causal variant due to linkage disequilibrium (LD). This paper addresses the challenge of accurately identifying which coding variants are truly causal for T2D risk, moving beyond simple association to robust validation.
Methods
The study employed a two-pronged approach:
Discovery: They aggregated exome-array data from a massive dataset of over 81,000 T2D cases and 370,000 controls across diverse ancestries. This large sample size significantly boosted their power to detect coding variant associations, especially for variants with modest effects. They used a weighted Bonferroni correction to account for the enrichment of trait associations in coding regions.
Fine-mapping: For the identified association signals, they performed high-resolution fine-mapping using European ancestry GWAS data (50,160 T2D cases and 465,272 controls). This involved constructing "credible sets" of variants that collectively explain 99% of the posterior probability of driving the association. They used two models:
Functionally unweighted: This model treats all variants equally, relying solely on genetic data.
Annotation-informed: This model incorporates prior knowledge about the functional impact of variants (e.g., protein-truncating variants, missense variants) to "boost" their posterior probabilities. This approach leverages the observation that coding variants are generally enriched for causal roles in GWAS.
Datasets
The study utilized several large-scale datasets:
Discovery Phase: Exome array data from up to 452,244 individuals (effective sample size 228,825) across five ancestry groups, including data from UK Biobank, GERA, and the GoT2D/T2D-GENES and SIGMA studies.
Fine-mapping Phase: European ancestry GWAS data from 50,160 T2D cases and 465,272 controls, aggregated from 24 studies by the DIAGRAM Consortium. This data was imputed using various reference panels (Haplotype Reference Consortium, 1000 Genomes Project, UK10K Project, deCODE Icelandic reference panel).
Key Insights and Ideas
Coding variants are disproportionately important: Even though coding regions make up a small fraction of the genome, they contribute significantly to the heritability of complex traits like T2D.
Association doesn't always mean causation: A significant association signal for a coding variant doesn't automatically mean it's the causal one. Linkage disequilibrium can make it hard to distinguish the true driver from nearby variants.
Fine-mapping is essential for validation: Precisely mapping association signals to specific variants, especially using models that incorporate functional annotations, is critical for accurately identifying causal variants.
"False leads" are common: Naïve analysis of associated coding variants can lead to incorrect conclusions about the effector transcript and disease mechanism.
Key Results
They identified 40 distinct coding variant association signals (at 38 loci) reaching genome-wide significance.
Of these, 16 represent novel associations not previously identified by GWAS.
Using fine-mapping with an annotation-informed model, they found convincing support that coding variants are causal for 16 of the 37 signals analyzed. This includes novel signals involving genes like POC5, ANKH, WSCD2, PLCB3, and PNPLA3.
However, at 13 of the 37 loci, the associated coding variants were identified as "false leads," meaning they were unlikely to be the true causal variants.
The study also highlighted that most T2D association signals were driven by common variants with modest effects, with only a few signals attributable to lower-frequency variants.
Limitations and Future Work
Rare variants are underrepresented: The exome array used in the study has limited coverage of very rare variants, meaning that some potentially important rare variants may have been missed. Future studies with whole-genome sequencing and larger, more diverse sample sizes will be needed to fully explore the role of rare variants.
Imputation imperfections: While imputation methods have improved, imperfections can still affect fine-mapping precision at some loci, requiring careful interpretation.
Need for functional validation: While fine-mapping provides strong evidence, definitive confirmation of causality often requires experimental functional studies to demonstrate how a variant affects gene function and links to disease phenotype.
Improved functional annotation: The authors suggest that future improvements in functional annotation, especially for regulatory variants and across relevant tissues, will enable more granular and accurate fine-mapping. This will be crucial for prioritizing targets for drug development.
Summary updated
Abstract
Identification of coding variant associations for complex diseases offers a direct route to biological insight, but is dependent on appropriate inference concerning the causal impact of those variants on disease risk. We aggregated coding variant data for 81,412 type 2 diabetes (T2D) cases and 370,832 controls of diverse ancestry, identifying 40 distinct coding variant association signals (at 38 loci) reaching significance ( p <2.2×10 −7 ). Of these, 16 represent novel associations mapping outside known genome-wide association study (GWAS) signals. We make two important observations. First, despite a threefold increase in sample size over previous efforts, only five of the 40 signals are driven by variants with minor allele frequency <5%, and we find no evidence for low-frequency variants with allelic odds ratio >1.29. Second, we used GWAS data from 50,160 T2D cases and 465,272 controls of European ancestry to fine-map these associated coding variants in their regional context, with and without additional weighting to account for the global enrichment of complex trait association signals in coding exons. At the 37 signals for which we attempted fine-mapping, we demonstrate convincing support (posterior probability >80% under the “annotation-weighted” model) that coding variants are causal for the association at 16 (including novel signals involving POC5 p.His36Arg, ANKH p.Arg187Gln, WSCD2 p.Thr113Ile, PLCB3 p.Ser778Leu, and PNPLA3 p.Ile148Met). However, at 13 of the 37 loci, the associated coding variants represent “false leads” and naïve analysis could have led to an erroneous inference regarding the effector transcript mediating the signal. Accurate identification of validated targets is dependent on correct specification of the contribution of coding and non-coding mediated mechanisms at associated loci.
BibTeX
@misc{mahajan2017refining,
author = {Anubha Mahajan and Jennifer Wessel and Sara M. Willems and Zhao Wei and Neil R. Robertson and Audrey Y. Chu and Wei Gan and Hidetoshi Kitajima and Daniel Taliun and N. William Rayner and Xiuqing Guo and Yingchang Lu and Man Li and Richard A. Jensen and Yao Hu and Shaofeng Huo and Kurt K. Lohman and Weihua Zhang and James P. Cook and Bram P. Prins and Jason Flannick and Niels Grarup and Vassily Trubetskoy and Jasmina Kravić and Young Jin Kim and Denis Rybin and Hanieh Yaghootkar and Martina Müller‐Nurasyid and Karina Meidtner and Ruifang Li‐Gao and Tibor V. Varga and Jonathan Marten and Li Jin and Albert V. Smith and Ping An and Symen Ligthart and Stefan Gustafsson and Giovanni Malerba and Ayşe Demirkan and Juan Fernández Tajes and Valgerður Steinthórsdóttir and Matthias Wuttke and Cécile Lecœur and Michael Preuß and Lawrence F. Bielak and Marielisa Graff and Heather M. Highland and Anne E. Justice and Dajiang J. Liu and Eirini Marouli and Gina M. Peloso and Helen R. Warren and Saima Afaq and Shoaib Afzal and Emma Ahlqvist and Peter Almgren and Najaf Amin and Lia B. Bang and Alain G. Bertoni and Cristina Bombieri and Jette Bork‐Jensen and Ivan Brandslund and Jennifer A. Brody and Noël P. Burtt and Mickaël Canouil and Yii‐Der Ida Chen and Yoon Shin Cho and Cramer Christensen and Sophie V. Eastwood and Kai‐Uwe Eckardt and Krista Fischer and Giovanni Gambaro and Vilmantas Giedraitis and Megan L. Grove and Hugoline G. de Haan and Sophie Hackinger and Yang Hai and Sohee Han and Anne Tybjærg‐Hansen and Marie‐France Hivert and Bo Isomaa and Susanne Jäger and Marit E. Jørgensen and Torben Jørgensen and Annemari Käräjämäki and Bong-Jo Kim and Sung‐Soo Kim and Heikki A. Koistinen and Péter Kovács and Jennifer Kriebel and Florian Kronenberg and Kristi Läll and Leslie A. Lange and Jung‐Jin Lee and Benjamin Lehne and Huaixing Li and Keng‐Hung Lin and Allan Linneberg and Ching‐Ti Liu and Jun Liu},
title = {{Refining the accuracy of validated target identification through coding variant fine-mapping in type 2 diabetes}},
year = {2017},
doi = {10.1101/144410},
howpublished = {bioRxiv},
}