Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 93 of the 93 references it cites were found in a registry or an index.

Exome sequencing of 20,791 cases of type 2 diabetes and 24,440 controls

Open access
Broad Genomics Platform, Flannick, Jason, DiscovEHR Collaboration, CHARGE, LuCamp, ProDiGY, GoT2D, ESP, SIGMA-T2D, T2D-GENES, AMP-T2D-GENES, Mercader, Josep M., Fuchsberger, Christian, Udler, Miriam S., Mahajan, Anubha, Wessel, Jennifer, Teslovich, Tanya M., Caulkins, Lizz, Koesterer, Ryan, Barajas-Olmos, Francisco, Blackwell, Thomas W., Boerwinkle, Eric, Brody, Jennifer A., Centeno-Cruz, Federico, Chen, Ling, Chen, Siying, Contreras-Cubas, Cecilia, Córdova, Emilio, Correa, Adolfo, Cortes, Maria, DeFronzo, Ralph A., Dolan, Lawrence, Drews, Kimberly L., Elliott, Amanda, Floyd, James S., Gabriel, Stacey, Garay-Sevilla, Maria Eugenia, García-Ortiz, Humberto, Gross, Myron, Han, Sohee, Heard-Costa, Nancy L., Jackson, Anne U., Jørgensen, Marit E., Kang, Hyun Min, Kelsey, Megan, Kim, Bong-Jo, Koistinen, Heikki A., Kuusisto, Johanna, Leader, Joseph B., Linneberg, Allan, Liu, Ching-Ti, Liu, Jianjun, Lyssenko, Valeriya, Manning, Alisa K., Marcketta, Anthony, Malacara-Hernandez, Juan Manuel, Martínez-Hernández, Angélica, Matsuo, Karen, Mayer-Davis, Elizabeth, Mendoza-Caamal, Elvia, Mohlke, Karen L., Morrison, Alanna C., Ndungu, Anne, Ng, Maggie C. Y., O’Dushlaine, Colm, Payne, Anthony J., Pihoker, Catherine, Post, Wendy S., Preuss, Michael, Psaty, Bruce M., Vasan, Ramachandran S., Rayner, N. William, Reiner, Alexander P., Revilla-Monsalve, Cristina, Robertson, Neil R., Santoro, Nicola, Schurmann, Claudia, So, Wing Yee, Soberón, Xavier, Stringham, Heather M., Strom, Tim M., Tam, Claudia H. T., Thameem, Farook, Tomlinson, Brian, Torres, Jason M., Tracy, Russell P., van Dam, Rob M., Vujkovic, Marijana, Wang, Shuai, Welch, Ryan P., Witte, Daniel R., Wong, Tien-Yin, Atzmon, Gil, Barzilai, Nir, Blangero, John, Bonnycastle, Lori L., Bowden, Donald W., Chambers, John C., Chan, Edmund, Cheng, Ching-Yu, Cho, Yoon Shin, Collins, Francis S., de Vries, Paul S., Duggirala, Ravindranath, Glaser, Benjamin, Gonzalez, Clicerio, Gonzalez, Ma Elena, Groop, Leif, Kooner, Jaspal Singh, Kwak, Soo Heon, Laakso, Markku, Lehman, Donna M., Nilsson, Peter, Spector, Timothy D., Tai, E. Shyong, Tuomi, Tiinamaija, Tuomilehto, Jaakko, Wilson, James G., Aguilar-Salinas, Carlos A., Bottinger, Erwin, Burke, Brian, Carey, David J., Chan, Juliana C. N., Dupuis, Josée, Frossard, Philippe, Heckbert, Susan R., Hwang, Mi Yeong, Kim, Young Jin, Kirchner, H. Lester, Lee, Jong-Young, Lee, Juyoung, Loos, Ruth J. F., Ma, Ronald C. W., Morris, Andrew D., O’Donnell, Christopher J., Palmer, Colin N. A., Pankow, James, Park, Kyong Soo, Rasheed, Asif, Saleheen, Danish, Sim, Xueling, Small, Kerrin S., Teo, Yik Ying, Haiman, Christopher, Hanis, Craig L., Henderson, Brian E., Orozco, Lorena, Tusié-Luna, Teresa, Dewey, Frederick E., Baras, Aris, Gieger, Christian, Meitinger, Thomas, Strauch, Konstantin, Lange, Leslie, Grarup, Niels, Hansen, Torben, Pedersen, Oluf, Zeitler, Philip, Dabelea, Dana, Abecasis, Goncalo, Bell, Graeme I., Cox, Nancy J., Seielstad, Mark, Sladek, Rob, Meigs, James B., Rich, Steve S., Rotter, Jerome I., Altshuler, David, Burtt, Noël P., Scott, Laura J., Morris, Andrew P., Florez, Jose C., McCarthy, Mark I., Boehnke, Michael · 2019 · Nature

Abstract

Protein-coding genetic variants that strongly affect disease risk can yield relevant clues to disease pathogenesis. Here we report exome-sequencing analyses of 20,791 individuals with type 2 diabetes (T2D) and 24,440 non-diabetic control participants from 5 ancestries. We identify gene-level associations of rare variants (with minor allele frequencies of less than 0.5%) in 4 genes at exome-wide significance, including a series of more than 30 SLC30A8 alleles that conveys protection against T2D, and in 12 gene sets, including those corresponding to T2D drug targets (P = 6.1 × 10−3) and candidate genes from knockout mice (P = 5.2 × 10−3). Within our study, the strongest T2D gene-level signals for rare variants explain at most 25% of the heritability of the strongest common single-variant signals, and the gene-level effect sizes of the rare variants that we observed in established T2D drug targets will require 75,000–185,000 sequenced cases to achieve exome-wide significance. We propose a method to interpret these modest rare-variant associations and to incorporate these associations into future target or gene prioritization efforts. Exome-sequencing analyses of a large cohort of patients with type 2 diabetes and control individuals without diabetes from five ancestries are used to identify gene-level associations of rare variants that are associated with type 2 diabetes.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41586-019-1231-2

Cites

93 of 93 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Exome-wide association analysis. a, Single-variant associations were calculated using the two-sided EMMAX test. Red line, P = 4.3 × 10−7. b, Gene-level P values, corrected for four analyses performed using the two-sided minimum P-value test. The most-significant genes are labelled. Red line, P = 6.5 × 10−7. c, SLC30A8 gene-level P values (left y axis, black line), calculated using a two-sided burden test after removing variants (in order of increasing single-variant P value) from the 1/5 1% mask (the strongest signal for SLC30A8). Dashed line, P = 0.05. Right y axis (blue line), estimated effect size (log10(odds ratio)). Blue shading, 95% confidence interval. Dotted line, effect size = 0. Single-variant n = 45,231 individuals. Gene-level n = 43,074 unrelated individuals.
Fig. 1 Exome-wide association analysis. a, Single-variant associations were calculated using the two-sided EMMAX test. Red line, P = 4.3 × 10−7. b, Gene-level P values, corrected for four analyses performed using the two-sided minimum P-value test. The most-significant genes are labelled. Red line, P = 6.5 × 10−7. c, SLC30A8 gene-level P values (left y axis, black line), calculated using a two-sided burden test after removing variants (in order of increasing single-variant P value) from the 1/5 1% mask (the strongest signal for SLC30A8). Dashed line, P = 0.05. Right y axis (blue line), estimated effect size (log10(odds ratio)). Blue shading, 95% confidence interval. Dotted line, effect size = 0. Single-variant n = 45,231 individuals. Gene-level n = 43,074 unrelated individuals.CC BY · PubMed Central
Fig. 2 Gene set analysis. a–e, Rank percentiles (1 = highest) for gene-level associations (compared to matched genes) within 11 monogenic diabetes genes (548 matched genes) (a), 8 T2D drug targets (400 matched genes) (b), 31 genes linked to NIDD in mice (1,499 matched genes) (c), 323 genes linked to impaired glucose tolerance in mice (10,043 matched genes) (d) and 11 genes with common likely causal coding variants (537 matched genes) (e). P values are from a one-sided Wilcoxon rank-sum test between each gene set and comparison set. Labels indicate minimum, 25th percentile, median, 75th percentile and maximum. f, Estimated odds ratios, from the weighted burden test of the 5/5 mask, for 8 T2D drug targets (red, agonists; blue, inhibitors). Data are mean ± s.e. of log(odds ratio) from the burden test. n = 43,071 unrelated individuals.
Fig. 2 Gene set analysis. a–e, Rank percentiles (1 = highest) for gene-level associations (compared to matched genes) within 11 monogenic diabetes genes (548 matched genes) (a), 8 T2D drug targets (400 matched genes) (b), 31 genes linked to NIDD in mice (1,499 matched genes) (c), 323 genes linked to impaired glucose tolerance in mice (10,043 matched genes) (d) and 11 genes with common likely causal coding variants (537 matched genes) (e). P values are from a one-sided Wilcoxon rank-sum test between each gene set and comparison set. Labels indicate minimum, 25th percentile, median, 75th percentile and maximum. f, Estimated odds ratios, from the weighted burden test of the 5/5 mask, for 8 T2D drug targets (red, agonists; blue, inhibitors). Data are mean ± s.e. of log(odds ratio) from the burden test. n = 43,071 unrelated individuals.CC BY · PubMed Central
Fig. 3 Comparison of exome-sequencing to array-based GWAS. a, Single-variant associations, calculated by two-sided Firth logistic regression, from an array-based imputed GWAS of the subset (n = 34,529) of samples in the exome-sequencing analysis for which array data were available. Labels and axes as described in Fig. 1a. b, The observed liability variance explained (LVE) by the top 19 exome-sequencing gene-level associations (red) and the top 19 imputed GWAS single-variant associations (maximum of 1 per 250 kb; blue) and their ratio (black). Signals ranked by liability variance explained. c, Gene rank percentiles from exome-sequencing gene-level analysis (x axis) and a previous multi-ancestry T2D GWAS13 (y axis). Genes are from the mouse NIDD gene set (Fig. 2c).
Fig. 3 Comparison of exome-sequencing to array-based GWAS. a, Single-variant associations, calculated by two-sided Firth logistic regression, from an array-based imputed GWAS of the subset (n = 34,529) of samples in the exome-sequencing analysis for which array data were available. Labels and axes as described in Fig. 1a. b, The observed liability variance explained (LVE) by the top 19 exome-sequencing gene-level associations (red) and the top 19 imputed GWAS single-variant associations (maximum of 1 per 250 kb; blue) and their ratio (black). Signals ranked by liability variance explained. c, Gene rank percentiles from exome-sequencing gene-level analysis (x axis) and a previous multi-ancestry T2D GWAS13 (y axis). Genes are from the mouse NIDD gene set (Fig. 2c).CC BY · PubMed Central
Fig. 4 Decision support from exome-sequencing data. a, Estimated power, as a function of future sample size (x axis), to detect T2D gene-level associations (two-sided test at P = 6.25 × 10−7) with aggregate frequency and odds ratios equal to those estimated from our analysis of 8 T2D drug targets (Fig. 2f). b, A proposed workflow for using exome-sequencing data in gene characterization. Depending on the prior belief in the disease relevance of a gene, the cost of experimental characterization and the benefit of gene validation, a decision to conduct the experiment could be informed by the posterior probability of the disease relevance of the gene, as estimated from exome-sequencing association statistics (available through http://www.type2diabetesgenetics.org/).
Fig. 4 Decision support from exome-sequencing data. a, Estimated power, as a function of future sample size (x axis), to detect T2D gene-level associations (two-sided test at P = 6.25 × 10−7) with aggregate frequency and odds ratios equal to those estimated from our analysis of 8 T2D drug targets (Fig. 2f). b, A proposed workflow for using exome-sequencing data in gene characterization. Depending on the prior belief in the disease relevance of a gene, the cost of experimental characterization and the benefit of gene validation, a decision to conduct the experiment could be informed by the posterior probability of the disease relevance of the gene, as estimated from exome-sequencing association statistics (available through http://www.type2diabetesgenetics.org/).CC BY · PubMed Central
Extended Data Fig. 1 Power analysis. The power to detect associations (using a two-sided test) at P < 5 × 10−8 for variants (or collections of variants) with a given minor allele frequency (x axis) and odds ratio (y axis) measured as the average across all ancestries. a, Cells are shaded according to the power of the current study of 20,791 T2D cases and 24,440 controls, with white indicating high power and red indicating low power. The 20%, 50%, 80% and 99% contour lines are labelled. b, Cells are shaded according to the difference in power between the current study and a previously published study of 12,940 individuals10, with yellow–white indicating a large increase in power and red indicating a small increase in power. The 20%, 40%, 60% and 80% contour lines are labelled.
Extended Data Fig. 1 Power analysis. The power to detect associations (using a two-sided test) at P < 5 × 10−8 for variants (or collections of variants) with a given minor allele frequency (x axis) and odds ratio (y axis) measured as the average across all ancestries. a, Cells are shaded according to the power of the current study of 20,791 T2D cases and 24,440 controls, with white indicating high power and red indicating low power. The 20%, 50%, 80% and 99% contour lines are labelled. b, Cells are shaded according to the difference in power between the current study and a previously published study of 12,940 individuals10, with yellow–white indicating a large increase in power and red indicating a small increase in power. The 20%, 40%, 60% and 80% contour lines are labelled.CC BY · PubMed Central
Extended Data Fig. 2 Data quality control workflow. A schematic of the steps involved in sample and variant quality control is shown. Quality control was conducted as described in the Methods to construct a set of samples and variants included in the association analysis. Each step is depicted as an arrow, with the number of samples or variants excluded by the step shown at the end of the arrow. The final set of samples and variants analysed are represented by the ‘Analysis’ dataset; we further excluded samples of high relatedness to other samples in the dataset from some but not all analyses. After each step that removed samples, we also removed newly monomorphic variants (hence the decrease in variants between the ‘Clean’ and ‘Analysis’ datasets).
Extended Data Fig. 2 Data quality control workflow. A schematic of the steps involved in sample and variant quality control is shown. Quality control was conducted as described in the Methods to construct a set of samples and variants included in the association analysis. Each step is depicted as an arrow, with the number of samples or variants excluded by the step shown at the end of the arrow. The final set of samples and variants analysed are represented by the ‘Analysis’ dataset; we further excluded samples of high relatedness to other samples in the dataset from some but not all analyses. After each step that removed samples, we also removed newly monomorphic variants (hence the decrease in variants between the ‘Clean’ and ‘Analysis’ datasets).CC BY · PubMed Central
Extended Data Fig. 3 Single-variant association analysis workflow. A schematic of the steps involved in single-variant exome-sequencing association analysis is shown, as described in the Methods. We began analysis with a division of samples in the ‘Analysis’ dataset (leftmost column) into 25 different subgroups (second column from the left) based on cohort, ancestry and sequencing technology. We then filtered variants according to metrics computed separately for each subgroup; we applied the filters listed in the ‘Basic filters’ box to all subgroups and for some subgroups we applied additional (more stringent) filters as indicated by boxes in the third column from the left. The resulting number of variants and samples advanced for analysis in each subgroup are indicated in the fourth column from the left. We analysed each subgroup with both the EMMAX test (to measure association strength) and the Firth test (to measure allelic odds ratios), each of which are two-sided; the number of principal components included as covariates in the Firth test is shown in the fifth column from the left. Finally, we combined each of the EMMAX and Firth subgroup-level results using a 25-group meta-analysis to produce the final P values and odds ratios reported for each variant. Multi, variant is multiallelic; CR, call rate; P, variant subgroup-level P value; P(Fisher), variant subgroup-level P value from Fisher’s exact test; P(miss), P value for subgroup-level variant differential missingness between T2D cases and controls; P(HWE), P value for deviation from subgroup-level Hardy–Weinberg equilibrium; Alt GQ, mean genotype quality of non-reference genotypes (across all samples); X Chrom, variant is on X chromosome.
Extended Data Fig. 3 Single-variant association analysis workflow. A schematic of the steps involved in single-variant exome-sequencing association analysis is shown, as described in the Methods. We began analysis with a division of samples in the ‘Analysis’ dataset (leftmost column) into 25 different subgroups (second column from the left) based on cohort, ancestry and sequencing technology. We then filtered variants according to metrics computed separately for each subgroup; we applied the filters listed in the ‘Basic filters’ box to all subgroups and for some subgroups we applied additional (more stringent) filters as indicated by boxes in the third column from the left. The resulting number of variants and samples advanced for analysis in each subgroup are indicated in the fourth column from the left. We analysed each subgroup with both the EMMAX test (to measure association strength) and the Firth test (to measure allelic odds ratios), each of which are two-sided; the number of principal components included as covariates in the Firth test is shown in the fifth column from the left. Finally, we combined each of the EMMAX and Firth subgroup-level results using a 25-group meta-analysis to produce the final P values and odds ratios reported for each variant. Multi, variant is multiallelic; CR, call rate; P, variant subgroup-level P value; P(Fisher), variant subgroup-level P value from Fisher’s exact test; P(miss), P value for subgroup-level variant differential missingness between T2D cases and controls; P(HWE), P value for deviation from subgroup-level Hardy–Weinberg equilibrium; Alt GQ, mean genotype quality of non-reference genotypes (across all samples); X Chrom, variant is on X chromosome.CC BY · PubMed Central
Extended Data Fig. 4 Calibration of single-variant analysis. To assess whether our single-variant association statistics (two-sided, calculated by the EMMAX test) were well-calibrated, we computed quantile–quantile plots of associations across all samples (Overall) and within each ancestry (total n = 45,231 individuals). To avoid deflation of the quantile–quantile plot from rare variants (for which the expected P values are discrete rather than uniformly distributed), only variants with minor allele counts of 20 or greater (either overall or within the relevant ancestry) are shown. Variants were also LD-pruned before plotting, to avoid induced variance from correlated P values of these variants, using the ‘clump’ method implemented in PLINK. The λ values indicate genomic control, as measured by the ratio in observed median χ2 statistic to that expected under the null hypothesis. Red line, expectation of P values under the null distribution. Blue lines (and grey region), 95% confidence interval of expectations under the null distribution.
Extended Data Fig. 4 Calibration of single-variant analysis. To assess whether our single-variant association statistics (two-sided, calculated by the EMMAX test) were well-calibrated, we computed quantile–quantile plots of associations across all samples (Overall) and within each ancestry (total n = 45,231 individuals). To avoid deflation of the quantile–quantile plot from rare variants (for which the expected P values are discrete rather than uniformly distributed), only variants with minor allele counts of 20 or greater (either overall or within the relevant ancestry) are shown. Variants were also LD-pruned before plotting, to avoid induced variance from correlated P values of these variants, using the ‘clump’ method implemented in PLINK. The λ values indicate genomic control, as measured by the ratio in observed median χ2 statistic to that expected under the null hypothesis. Red line, expectation of P values under the null distribution. Blue lines (and grey region), 95% confidence interval of expectations under the null distribution.CC BY · PubMed Central
Extended Data Fig. 5 Gene-level association analysis workflow. A schematic of the steps involved in gene-level exome-sequencing association analysis, as described in Methods, is shown. We began analysis with subgroup-level genotype filtering (second column from the left) of unrelated samples in the ‘Analysis’ dataset (leftmost column); we then applied genotype filters for each subgroup (filtering genotypes for either all or no samples in each subgroup), similar to those used in subgroup-level single-variant analyses. We then annotated each non-reference variant allele with 16 different bioinformatics algorithms to assess allele deleteriousness, and we grouped alleles into one of seven nested masks (third column from the left; the number of variants and weights shown correspond to alleles absent from ‘higher’, or more stringent, nested masks). We computed burden and SKAT analyses (both of which are two-sided) using one of two approaches to combine alleles across masks (Methods): first, by analysing all alleles at once with weights assigned according to the most stringent mask containing the allele (weighted test); and second, by analysing each mask independently and then calculating the lowest P value corrected for the effective number of tests (minimum P-value test). Multi, variant is multiallelic; CR, call rate; P(miss), P value for subgroup-level variant differential missingness between T2D cases and controls; P(HWE), P value for deviation from subgroup-level Hardy–Weinberg equilibrium; Alt GQ, mean genotype quality of non-reference genotypes (across all samples).
Extended Data Fig. 5 Gene-level association analysis workflow. A schematic of the steps involved in gene-level exome-sequencing association analysis, as described in Methods, is shown. We began analysis with subgroup-level genotype filtering (second column from the left) of unrelated samples in the ‘Analysis’ dataset (leftmost column); we then applied genotype filters for each subgroup (filtering genotypes for either all or no samples in each subgroup), similar to those used in subgroup-level single-variant analyses. We then annotated each non-reference variant allele with 16 different bioinformatics algorithms to assess allele deleteriousness, and we grouped alleles into one of seven nested masks (third column from the left; the number of variants and weights shown correspond to alleles absent from ‘higher’, or more stringent, nested masks). We computed burden and SKAT analyses (both of which are two-sided) using one of two approaches to combine alleles across masks (Methods): first, by analysing all alleles at once with weights assigned according to the most stringent mask containing the allele (weighted test); and second, by analysing each mask independently and then calculating the lowest P value corrected for the effective number of tests (minimum P-value test). Multi, variant is multiallelic; CR, call rate; P(miss), P value for subgroup-level variant differential missingness between T2D cases and controls; P(HWE), P value for deviation from subgroup-level Hardy–Weinberg equilibrium; Alt GQ, mean genotype quality of non-reference genotypes (across all samples).CC BY · PubMed Central
Extended Data Fig. 6 Calibration of gene-level association analyses. For both the burden and SKAT tests, we tested for gene-level association within seven different allelic masks. As this produced seven P values for each test, we developed two means to consolidate these results (Methods). a, b, The quantile–quantile plots of associations are shown for the minimum P-value burden test (a) and the weighted burden test (b). Each test is two-sided and consists of n = 43,071 unrelated individuals. Only genes with combined minor allele count of 20 or greater are shown in the quantile–quantile plots, to avoid deflation from genes with too few variants to produce P values asymptotically uniform under the null distribution. The λ values indicate genomic control, as measured by the ratio in observed median χ2 statistic to that expected under the null hypothesis. The three genes with exome-wide significant associations are labelled. Red line, expectation of P values under the null distribution. Blue lines (and grey region), 95% confidence interval of expectations under the null distribution.
Extended Data Fig. 6 Calibration of gene-level association analyses. For both the burden and SKAT tests, we tested for gene-level association within seven different allelic masks. As this produced seven P values for each test, we developed two means to consolidate these results (Methods). a, b, The quantile–quantile plots of associations are shown for the minimum P-value burden test (a) and the weighted burden test (b). Each test is two-sided and consists of n = 43,071 unrelated individuals. Only genes with combined minor allele count of 20 or greater are shown in the quantile–quantile plots, to avoid deflation from genes with too few variants to produce P values asymptotically uniform under the null distribution. The λ values indicate genomic control, as measured by the ratio in observed median χ2 statistic to that expected under the null hypothesis. The three genes with exome-wide significant associations are labelled. Red line, expectation of P values under the null distribution. Blue lines (and grey region), 95% confidence interval of expectations under the null distribution.CC BY · PubMed Central
Extended Data Fig. 7 PPA calculation workflow. a, We estimated the PPAs for nonsynonymous variants in our sequence analysis based on concordance with independent exome array data and previously published6,78,80 estimates of the fraction of causal coding associations (Methods). b, PPA estimates for nonsynonymous variants within T2D GWAS loci are shown as a function of P value (right y axis, black line; 95% confidence interval, grey shading) together with the total number of such variants (left y axis, red line). c, For variants outside of T2D GWAS loci, we developed a method to further compute Bayes factors, which measure the odds of true and causal association, as a function of P value, using a model of the prior odds of true and causal association for variants in GWAS loci (Methods). d, These Bayes factors can be combined with a subjective prior belief in the T2D relevance of a gene (y axis) to produce the estimated posterior probability of true and causal association for any nonsynonymous variant in the exome-sequence dataset based on its observed log10(P) (x axis). Posterior estimates are shaded proportional to value (red, low; white, high). Values are shown for the default modelling assumptions of 33% of missense variants that caused gene inactivation and 30% of true missense associations that represented the causal variant.
Extended Data Fig. 7 PPA calculation workflow. a, We estimated the PPAs for nonsynonymous variants in our sequence analysis based on concordance with independent exome array data and previously published6,78,80 estimates of the fraction of causal coding associations (Methods). b, PPA estimates for nonsynonymous variants within T2D GWAS loci are shown as a function of P value (right y axis, black line; 95% confidence interval, grey shading) together with the total number of such variants (left y axis, red line). c, For variants outside of T2D GWAS loci, we developed a method to further compute Bayes factors, which measure the odds of true and causal association, as a function of P value, using a model of the prior odds of true and causal association for variants in GWAS loci (Methods). d, These Bayes factors can be combined with a subjective prior belief in the T2D relevance of a gene (y axis) to produce the estimated posterior probability of true and causal association for any nonsynonymous variant in the exome-sequence dataset based on its observed log10(P) (x axis). Posterior estimates are shaded proportional to value (red, low; white, high). Values are shown for the default modelling assumptions of 33% of missense variants that caused gene inactivation and 30% of true missense associations that represented the causal variant.CC BY · PubMed Central
Extended Data Fig. 8 Estimated posterior probability of associations for different prior hypotheses. We estimated the posterior probability of association for nonsynonymous variants that met various single-variant P-value thresholds (two-sided EMMAX test, n = 45,231 individuals) in our analysis, as described in the Methods and shown in Extended Data Fig. 7. To perform the needed calculations, we assumed that, on average, 1.1 genes that are found within each T2D GWAS locus are relevant to T2D and 33% of missense mutations within these genes cause gene loss-of-function. a–f, To assess the sensitivity of our analysis to these assumptions, we repeated the calculations with different assumptions of 0.5 (a), 2.0 (b), 0.25 (c) and 0.1 (d) T2D-relevant genes within each GWAS locus, as well as 25% (e) and 40% (f) of missense variants leading to loss-of-function. All analyses assume the default modelling parameters that 30% of true nonsynonymous associations are causal associations; different values for this parameter would scale posterior probability estimates linearly.
Extended Data Fig. 8 Estimated posterior probability of associations for different prior hypotheses. We estimated the posterior probability of association for nonsynonymous variants that met various single-variant P-value thresholds (two-sided EMMAX test, n = 45,231 individuals) in our analysis, as described in the Methods and shown in Extended Data Fig. 7. To perform the needed calculations, we assumed that, on average, 1.1 genes that are found within each T2D GWAS locus are relevant to T2D and 33% of missense mutations within these genes cause gene loss-of-function. a–f, To assess the sensitivity of our analysis to these assumptions, we repeated the calculations with different assumptions of 0.5 (a), 2.0 (b), 0.25 (c) and 0.1 (d) T2D-relevant genes within each GWAS locus, as well as 25% (e) and 40% (f) of missense variants leading to loss-of-function. All analyses assume the default modelling parameters that 30% of true nonsynonymous associations are causal associations; different values for this parameter would scale posterior probability estimates linearly.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 86 references

Cited by

See all 328

Metadata

Authors
Platform, B. G., Flannick, J., DiscovEHR Collaboration, CHARGE, LuCamp, ProDiGY, GoT2D, ESP, SIGMA-T2D, T2D-GENES, AMP-T2D-GENES, Mercader, J. M., Fuchsberger, C., Udler, M. S., Mahajan, A., Wessel, J., Teslovich, T. M., Caulkins, L., Koesterer, R., Barajas-Olmos, F., Blackwell, T. W., Boerwinkle, E., Brody, J. A., Centeno-Cruz, F., Chen, L., Chen, S., Contreras-Cubas, C., Córdova, E., Correa, A., Cortes, M., DeFronzo, R. A., Dolan, L., Drews, K. L., Elliott, A., Floyd, J. S., Gabriel, S., Garay-Sevilla, M. E., García-Ortiz, H., Gross, M., Han, S., Heard-Costa, N. L., Jackson, A. U., Jørgensen, M. E., Kang, H. M., Kelsey, M., Kim, B.-J., Koistinen, H. A., Kuusisto, J., Leader, J. B., Linneberg, A., Liu, C.-T., Liu, J., Lyssenko, V., Manning, A. K., Marcketta, A., Malacara-Hernandez, J. M., Martínez-Hernández, A., Matsuo, K., Mayer-Davis, E., Mendoza-Caamal, E., Mohlke, K. L., Morrison, A. C., Ndungu, A., Ng, M. C. Y., O’Dushlaine, C., Payne, A. J., Pihoker, C., Post, W. S., Preuss, M., Psaty, B. M., Vasan, R. S., Rayner, N. W., Reiner, A. P., Revilla-Monsalve, C., Robertson, N. R., Santoro, N., Schurmann, C., So, W. Y., Soberón, X., Stringham, H. M., Strom, T. M., Tam, C. H. T., Thameem, F., Tomlinson, B., Torres, J. M., Tracy, R. P., van Dam, R. M., Vujkovic, M., Wang, S., Welch, R. P., Witte, D. R., Wong, T.-Y., Atzmon, G., Barzilai, N., Blangero, J., Bonnycastle, L. L., Bowden, D. W., Chambers, J. C., Chan, E., Cheng, C.-Y., Cho, Y. S., Collins, F. S., de Vries, P. S., Duggirala, R., Glaser, B., Gonzalez, C., Gonzalez, M. E., Groop, L., Kooner, J. S., Kwak, S. H., Laakso, M., Lehman, D. M., Nilsson, P., Spector, T. D., Tai, E. S., Tuomi, T., Tuomilehto, J., Wilson, J. G., Aguilar-Salinas, C. A., Bottinger, E., Burke, B., Carey, D. J., Chan, J. C. N., Dupuis, J., Frossard, P., Heckbert, S. R., Hwang, M. Y., Kim, Y. J., Kirchner, H. L., Lee, J.-Y., Lee, J., Loos, R. J. F., Ma, R. C. W., Morris, A. D., O’Donnell, C. J., Palmer, C. N. A., Pankow, J., Park, K. S., Rasheed, A., Saleheen, D., Sim, X., Small, K. S., Teo, Y. Y., Haiman, C., Hanis, C. L., Henderson, B. E., Orozco, L., Tusié-Luna, T., Dewey, F. E., Baras, A., Gieger, C., Meitinger, T., Strauch, K., Lange, L., Grarup, N., Hansen, T., Pedersen, O., Zeitler, P., Dabelea, D., Abecasis, G., Bell, G. I., Cox, N. J., Seielstad, M., Sladek, R., Meigs, J. B., Rich, S. S., Rotter, J. I., Altshuler, D., Burtt, N. P., Scott, L. J., Morris, A. P., Florez, J. C., McCarthy, M. I., & Boehnke, M.
Affiliations
  • Broad Institute (US)
  • Eurac Research (IT)
  • Massachusetts General Hospital (US)
  • Centre for Human Genetics (GB)
  • Indiana University Indianapolis (US)
  • Regeneron (United States) (US)
  • National Institute of Genomic Medicine (MX)
  • Harvard University (US)
  • University of Mississippi Medical Center (US)
  • The University of Texas at San Antonio Health Science Center (US)
  • Cincinnati Children's Hospital Medical Center (US)
  • George Washington University (US)
  • University of Washington (US)
  • University of Chicago (US)
  • Boston University (US)
  • University of Minnesota (US)
  • Korea National Institute of Health (KR)
  • Agency for Science, Technology and Research (SG)
  • Lund University (SE)
  • University of Helsinki (FI)
  • University of Eastern Finland (FI)
  • University of Michigan (US)
  • University of North Carolina at Chapel Hill (US)
  • Seattle Children's Hospital (US)
  • Framingham Heart Study (US)
  • Mexican Social Security Institute (MX)
  • Johns Hopkins University (US)
  • Chinese University of Hong Kong (HK)
  • Kaiser Permanente Washington Health Research Institute (US)
  • University of Vermont (US)
  • National University of Singapore (SG)
  • Albert Einstein College of Medicine (US)
  • Aarhus University (DK)
  • Hallym University (KR)
  • National Institutes of Health (US)
  • The University of Texas Rio Grande Valley (US)
  • Instituto Nacional de Salud Pública (MX)
  • Imperial College Healthcare NHS Trust (GB)
  • Duke-NUS Medical School (SG)
  • Seoul National University (KR)
  • Universität für Weiterbildung Krems (AT)
  • Chungbuk National University (KR)
  • Ninewells Hospital (GB)
  • Brigham and Women's Hospital (US)
  • Imperial College London (GB)
  • University of Southern California (US)
  • Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (MX)
  • King's College London (GB)
  • The University of Texas Health Science Center at Houston (US)
  • University of Copenhagen (DK)
  • Colorado School of Public Health (US)
  • Helmholtz Munich (DE)
  • Vanderbilt University (US)
  • University of California, San Francisco (US)
  • University of California, Los Angeles (US)
  • University of Liverpool (GB)
Corresponding author
Flannick, Jason
ORCID
Venue
Nature
Published
2019-05-22
Volume
570
Issue
7759
Pages
71-76
ISSN
  • 0028-0836 (ISSN-L)
  • 1476-4687
OpenAlex
W2945877977
Cited by
304
References
86
Citations by year
2026: 16 · 2025: 22 · 2024: 38 · 2023: 29 · 2022: 64 · 2021: 67 · 2020: 46 · 2019: 21
FWCI
31.04
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, cc-by)
  • Research Explorer (The University of Manchester) (publishedVersion, other-oa)
  • University of Southern Denmark Research Portal (University of Southern Denmark) (acceptedVersion, cc-by)
  • Työväentutkimus Vuosikirja (submittedVersion)
  • PuSH - Publication Server of Helmholtz Zentrum München (submittedVersion, cc-by)
  • Digital Access to Scholarship at Harvard (DASH) (Harvard University) (submittedVersion)
  • ScholarWorks @ UTRGV (The University of Texas Rio Grande Valley) (submittedVersion, cc-by)
  • Discovery Research Portal (University of Dundee) (submittedVersion, cc-by)
  • eScholarship (California Digital Library) (submittedVersion, cc-by)
  • Archivio Istituzionale della Ricerca - Università degli Studi del Molise (submittedVersion)
  • Minerva Access (University of Melbourne) (submittedVersion, cc-by)
  • Research Portal (King's College London) (submittedVersion)
  • Lund University Publications (Lund University) (submittedVersion)
  • mediaTUM (Technical University of Munich) (submittedVersion, other-oa)
  • mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) (submittedVersion, other-oa)
  • mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) (submittedVersion)
  • Spiral (Imperial College London) (submittedVersion, cc-by)
  • Open Research (University of Surrey) (submittedVersion, other-oa)
Funders
  • Massachusetts General Hospital (M01-RR00043-45, UL1-RR024139, UL1-RR024153, M01-RR00069, UL1-RR024992, UL1-RR024134, UL1-RR025780, M01RR01066, UL1-RR025758, UL1-RR024989, M01-RR00125, M01-RR14467, M01-RR00036, M01-RR00084)
  • Yale University (UL1-RR025780, M01-RR00084, M01-RR00036, M01-RR00069, UL1‐RR‐024153, M01-RR14467, UL1-RR024989, M01-RR00125, UL1-RR025758, M01-RR01066, UL1-RR024139, UL1-RR024134, M01-RR00043-45, UL1-RR024992)
  • Children's Hospital of Philadelphia (M01-RR00125, UL1-RR025780, UL1-RR-024134, M01-RR00084, M01-RR00043-45, M01-RR01066, M01-RR14467, UL1-RR025758, UL1-RR024992, M01-RR00036, UL1-RR024989, M01-RR00069, UL1-RR024139, UL1-RR024153)
  • Children's Hospital Los Angeles (UL1-RR025758, UL1-RR024134, UL1-RR024989, UL1-RR024153, M01-RR14467, M01-RR00036, UL1-RR024139, UL1-RR024992, M01-RR00084, M01-RR00043-45, M01-RR00069, M01-RR01066, UL1-RR025780, M01-RR00125)
  • Washington University in St. Louis (M01-RR14467, M01-RR00036, UL1-RR024139, UL1-RR025780, M01-RR01066, UL1-RR024134, UL1-RR025758, M01-RR00084, UL1-RR024989, UL1-RR024992, M01-RR00125, M01-RR00043-45, M01-RR00069, UL1-RR024153)
  • University of Oklahoma (M01-RR14467)
  • Case Western Reserve University (M01-RR01066, UL1RR024989, M01-RR14467, M01-RR00069, UL1-RR024992, M01-RR00084, M01-RR00036, M01-RR00043-45, UL1-RR024139, UL1RR024134, M01-RR00125, UL1-RR025758, UL1-RR025780, UL1-RR024153)
  • Wellcome Trust (095552/Z/11/Z, 092447, 090367, 085475, 212945/Z/18/Z, 064890, WT098017, 086596, WT098381, 090532, 083948)
  • Boston Area Diabetes Endocrinology Research Center (P30DK057521, DK057521)
  • Broad Institute (U01HG007417, U54HG003067)
  • Steno Diabetes Center Copenhagen (SDCC 3. B Epidemiology)
  • National Institute for Health and Care Research (NF-SI-0617-10090)
  • National Research Foundation
  • Ministerio de Economía y Competitividad
  • Lundbeckfonden (R16-2007-1691)
  • National Research Foundation of Korea (NRF-2017R1A2B4006508)
  • Chinese University of Hong Kong
  • Shanghai Jiao Tong University
  • Novo Nordisk Fonden (NNF18OC0034440)
  • Innovation and Technology Fund (ITS/088/08, ITS/487/09FP)
  • Hong Kong Government
  • National Institutes of Health (n01hc95164-17-0-1, n01hc085082-026, n01hc085081-021, n01hc085079-024, UL1TR001420, n01hc085083-027, n01hc95165-34-0-1, n01hc085080-026, n01hc95161-24-0-1, n01hc095169-003, n01hc95166-21-0-1, n01hc085086-022, n01hc95162-31-0-1, n01hc95168-27-0-1, n01hc95160-18-0-1, n01hc095167-000, 4845-301, 5rc2dk088389-02, UL1TR001079, n01hc25195-41-0-1, n01hc95163-22-0-1, 1rc2hl103010-01, n01hc55222-0-0-0, n01hc95159-40-0-1, UL1TR002548, DK020595, DK057521, RR‐00036, RR024153, RR025780, RR00069, M01-RR00043-45, RR024992, RR024989, UL1TR002345, UL1TR001863, UL1TR001881, 5r01hl113323-05)
  • Centers for Disease Control and Prevention
  • University of Oklahoma Health Sciences Center (UL1-RR024989, M01-RR00125, M01-RR00069, UL1-RR025780, UL1-RR024134, UL1-RR024139, M01-RR00043-45, M01-RR01066, M01-RR00036, UL1-RR025758, M01-RR00084, UL1-RR024992, UL1-RR024153, M01-RR14467)
  • University of Colorado Denver (UL1-RR025780, M01-RR01066, M01-RR00069, UL1-RR024153, M01-RR00084, UL1-RR024992, UL1-RR025758, UL1-RR024139, M01-RR00043-45, UL1-RR024134, M01-RR00125, M01-RR00036, UL1-RR024989, M01-RR14467)
  • Medical Research Council (MC_PC_13040)
  • National Medical Research Council (NMRC/CSASI/0012/2017, CSA-SI/0012/2017)
  • Korea Centers for Disease Control and Prevention (4851-302, 4845-301, 4851-307)
  • National Institute on Aging (R01AG042188, P01AG027734, P30AG038072, R01AG057909, R01AG023629, R01AG033193, R01AG046949, R01AG008122)
  • National Heart, Lung, and Blood Institute (N01 HC085086, RC2HL102925, N01-HC095169, N01-HC-095165, N01 HC085079, N01 HC095163, RC2HL102926, N01 HC085083, RC2HL102923, N01 HC085080, R01-HL113323, U01HL130114, RC2HL-103010, N01-HC-095159, N01 HC095168, N01-HC-095162, N01-HC-095161, N01-HC-025195, K01HL125751, N01 HC095167, N01HC085082, R01HL105756, U01HL080295, N01-HC-095160, N01 HC055222, N01 HC095166, N01 HC095164, RC2HL102924, N01 HC085081, RC2HL102419, R01HL092577, R01HL086694)
  • National Human Genome Research Institute (U54HG003067, U01HG007417, U54HG003273)
  • NIH Office of the Director (UL1-RR024153, M01-RR00069, UL1-RR025758, M01-RR01066, UL1-RR025780, UL1-RR024989, M01-RR00125, M01-RR00036, M01-RR00043-45, UL1-RR024139, UL1-RR024134, P30 DK020595, UL1-RR024992, M01-RR14467, M01-RR00084)
  • National Institute of General Medical Sciences (U54GM115428)
  • National Institute of Diabetes and Digestive and Kidney Diseases (U01DK085526, P30DK057521, U01DK061239, U01DK061212, R01DK114504, U01DK057295, K01DK107836, M01RR01066, R01DK072193, UL1-RR-024139, UL1-RR024989, U01DK062370, U01DK085524, R01DK066358, M01-RR00043-45, RC2-DK088389, DK020595, M01-RR00125, U01DK061254, U01DK085545, K24DK110550, R56DK062370, R01DK047482, M01-RR00036, U01DK061242, DK110550, U01DK061230, U01DK105535, K24DK080140, R01DK107786, DK-057521, R01DK110113, U01DK078616, R01DK101855, UL1-RR024992, R01DK053889, L30 DK106874, K23DK114551, M01-RR00069, M01-RR00084, R01DK098032, M01-RR14467, UL1-RR025758, R01DK093757, UM1DK105554, P30DK063491, P30DK020595, R01DK062370, U01DK105554)
  • National Center for Research Resources (RR-025780, M01RR000125, RR 024989, M01 RR-14467, M01 RR-00125, UL1RR024989, RR-024153, RR 00069, M01-RR00036, M01-RR-00084, M01 RR00069, UL1RR025758, M01RR001066, UL1RR024153, RR-024992, RR 024134, UL1RR024134, M01RR000043, M01‐RR00043‐45, M01RR014467, M01RR000069, RR024139, UL1RR024992, UL1RR024139, M01RR000084, M01-RR01066, RR.00036, M01RR000036, UL1RR025780)
  • National Center for Advancing Translational Sciences
  • Children's Hospital of Pittsburgh (UL1‐RR‐024989, UL1-RR025780, UL1-RR024153, M01-RR14467, M01-RR01066, M01-RR00043-45, UL1-RR024139, M01-RR00036, UL1-RR024992, M01-RR00125, UL1-RR024134, M01-RR00084, M01-RR00069, UL1-RR025758)
  • Steno Diabetes Center Aarhus
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genomics and Rare Diseases
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Diabetes and associated disorders
Keywords
type 2 diabetes, rare variants, SLC30A8, heritability, gene prioritization, ancestry
MeSH terms
Exome Sequencing, Animals, Decision Support Techniques, Diabetes Mellitus, Type 2, Female, Gene Frequency, Humans, Male, Case-Control Studies, Mice, Knockout, Mice, Genome-Wide Association Study, Exome

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Exome sequencing of 20,791 cases of type 2 diabetes and 24,440 controls. https://app.cochat.ai/papers/10.1038/s41586-019-1231-2
The paper
Platform, B. G., Flannick, J., DiscovEHR Collaboration, CHARGE, LuCamp, ProDiGY, GoT2D, ESP, SIGMA-T2D, T2D-GENES, AMP-T2D-GENES, Mercader, J. M., Fuchsberger, C., Udler, M. S., Mahajan, A., Wessel, J., Teslovich, T. M., Caulkins, L., Koesterer, R., . . . Boehnke, M. (2019). Exome sequencing of 20,791 cases of type 2 diabetes and 24,440 controls. Nature. https://doi.org/10.1038/s41586-019-1231-2

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.