The authors have declared that no competing interests exist.
¶ Members of the IIBDGC Consortium are listed in the Acknowledgments.
Genetic variants underlying complex traits, including disease susceptibility, are enriched within the transcriptional regulatory elements, promoters and enhancers. There is emerging evidence that regulatory elements associated with particular traits or diseases share similar patterns of transcriptional activity. Accordingly, shared transcriptional activity (coexpression) may help prioritise loci associated with a given trait, and help to identify underlying biological processes. Using cap analysis of gene expression (CAGE) profiles of promoter- and enhancer-derived RNAs across 1824 human samples, we have analysed coexpression of RNAs originating from trait-associated regulatory regions using a novel quantitative method (network density analysis; NDA). For most traits studied, phenotype-associated variants in regulatory regions were linked to tightly-coexpressed networks that are likely to share important functional characteristics. Coexpression provides a new signal, independent of phenotype association, to enable fine mapping of causative variants. The NDA coexpression approach identifies new genetic variants associated with specific traits, including an association between the regulation of the OCT1 cation transporter and genetic variants underlying circulating cholesterol levels. NDA strongly implicates particular cell types and tissues in disease pathogenesis. For example, distinct groupings of disease-associated regulatory regions implicate two distinct biological processes in the pathogenesis of ulcerative colitis; a further two separate processes are implicated in Crohn’s disease. Thus, our functional analysis of genetic predisposition to disease defines new distinct disease endotypes. We predict that patients with a preponderance of susceptibility variants in each group are likely to respond differently to pharmacological therapy. Together, these findings enable a deeper biological understanding of the causal basis of complex traits.
We discover that genetic variants associated with specific diseases have more in common with each other than we have previously seen. Specifically, variants associated with the same disease tend to be in parts of the genome that are turned on or off in similar complex patterns across many different cell types. We discover that genetic variants associated with specific diseases are found within regulatory elements that share patterns of expression. Specifically, variants associated with the same disease tend to be in parts of the genome that are turned on or off together in similar complex patterns across many different cell types. Knowing this helps us to find new variants associated with some diseases, and to better understand the genetic causes of other diseases. Furthermore, we discover that the genetic causes of inflammatory bowel disease fall into two distinct patterns, indicating that two aetiologically-distinct endotypes of this condition exist. Unlike other methods to learn about disease mechanisms from genetic information, our approach does not require any knowledge or assumptions about the genes themselves–it depends only on the patterns in which parts of the genome are activated in different cell types.
Genome-wide association studies (GWAS) have considerable untapped potential to reveal new mechanisms of disease[
We have recently shown that cell-type specific patterns of activity at multiple alternative promoters[
Unlike analysis of chromatin modifications or accessibility, the CAGE sequencing used in FANTOM5 combines extremely high resolution in three relevant dimensions: maximal spatial resolution on the genome, quantification of activity (transcript expression) over a wide dynamic range, and high biological resolution–quantifying activity in a much wider range of cell types and conditions than any previous study of regulatory variation[
Genes that are coexpressed are more likely to share common biology[
In contrast to previous studies[
In order to determine whether coexpression of regulatory elements can provide additional information to prioritise genome-wide associations that would otherwise fall below genome-wide significance, we developed network density analysis (NDA). The NDA method combines genetic signals (disease association in a GWAS) with functional signals (correlation in promoter and enhancer-associated transcript levels measured by CAGE across numerous cell types and tissues,
(a) A subset of regulatory elements is identified containing disease-associated SNPs. (b) The strength of the links between pairs of these regulatory regions is quantified, first as the Spearman correlation, then as the –
For the purpose of this analysis, promoters identified in the FANTOM5 dataset were defined as the region from -300 bases to +100 bases from a transcription start site[
For each GWAS study, SNPs were identified that lie within either a functional promoter or enhancer. Any promoter or enhancer that contained a variant putatively associated with a given phenotype was considered to be candidate phenotype-associated regulatory region. A pairwise matrix was then generated from the full FANTOM5 dataset of promoters and enhancers, in which each node is a regulatory region, and edges reflect the similarity in activity (expression) patterns arising at these regulatory regions, across different cell types and tissues.
To test the hypothesis that regulatory regions genetically associated with a given phenotype are more likely to share activity patterns, we devised the NDA method, which quantifies the strength of coexpression among a chosen pool of putative phenotype-associated regulatory regions. This approach avoids arbitrary cut-offs between clusters (or “communities”) of nodes, and yields a single value for each node, quantifying the closeness with all other nodes in a particular subset (network density). NDA was used to integrate the putative association between a regulatory sequence and the phenotype of interest (indicated by the presence of a phenotype-associated SNP), with the coexpression similarity between this node with other nodes that are also putatively associated with the same phenotype.
NDA integrates information from two distinct and independent sources: the relationships between nodes in the network, and the choice of subset. In the present work, nodes are regulatory regions, the subset is those regulatory regions that contain variants associated with a particular phenotype. Spearman’s rank correlation was chosen to quantify pairwise relationships, in view of the robustness of this measure in a variety of different distributions. However, the NDA approach is generalisable to any network of pairwise relationships.
Within a network of all possible pairwise relationships between nodes, a subset of nodes is selected that share a particular characteristic. Within this subset of nodes, every pair of nodes is considered. Each relationship between two nodes is expressed as the –
From the set of all nodes in a network, a subset is selected because they share some characteristic. In the case of the genomic analyses reported here, the nodes are TSS, and the subset of interest is those TSS that contain a variant that has some evidence of association with a particular trait. Throughout this paper, we have defined the set of phenotype-associated transcription start sites,
Input SNPs from GWAS results tend to be in LD with nearby variants. There is therefore a risk of spurious coexpression, since nearby regulatory regions are also likely to share regulatory influences, such as chromatin accessibility, enhancers, and lncRNAs. One solution to this would be to filter input SNPs by LD. However this would require that LD relationships for all SNPs be known for all of the populations from which SNP association data were derived, which is not the case. It would also risk removing functionally important regulatory regions from the analysis, by choosing only one SNP per LD block.
In order to overcome these problems, we sought to identify those regulatory region-associated SNPs within a given region that are most likely to contribute to a given subnetwork of putative phenotype-associated regulatory regions. By the definitions described above, these will be those regulatory regions with the highest NDA score. Regulatory regions are considered for combination if they are separated by 100,000bp or less. If any regulatory region within this range has a correlation
In order to confirm that spurious coexpression signals are not being generated solely because of LD, we used the ENSEMBL Perl API for the 1000 genomes phase 3 data (CEU) to search for variants in LD with each SNP lying within the chosen regulatory region for each group. Variants in LD with a variant in any other chosen regulatory region are reported.
For every node in the set
Raw
The node in the network with the highest NDA score has, by definition, numerous strong correlations with other nodes in the subset
Of 267,225 robust promoters and enhancers identified by FANTOM5, 93,558 (50.6%) were promoters within 400 bases of the 5′ end of a known transcript model. These were annotated with the name of the transcript. Alternative promoters were named in order of the highest transcriptional activity. Where necessary, coordinates for GWAS SNPs were translated to hg19 coordinates using LiftOver, or coordinates were obtained for SNP IDs from dbSNP version 138.
A circular permutation method was devised to prevent systematic bias by maintaining the underlying structure of GWAS SNP data. The NDA score for a given regulatory region was compared with NDA scores obtained from randomly permuted subsets of genes to give an empirical
Pre-mapping permutations use a random set of SNPs generated by rotation of the input set of SNPs,
In order to quantify the effect of coexpression alone (i.e. eliminating the inflation of NDA scores that occurs due to enrichment of trait-associated SNPs in regulatory regions), permuted networks were generated after mapping to TSS regions. This is analogous to randomly reassigning the labels in the network, but aims to preserve the local relationships between regulatory regions, since we cannot assume that regulatory regions are randomly distributed on the genome, and since regional regulatory events, such as chromatin reorganisation, are expected to lead to coexpression between nearby regulatory regions.
Where
This process generates a pool of variants that are likely to be grouped in a similar distribution on the genome to the input set. If the input set contains a large group of TSS regions in close proximity to each other on the genome, it is likely that this group of TSS regions will be joined as a single unit (see above) for analysis. During generation of permutations, the same number of consecutive TSS regions elsewhere on the genome may not be in sufficient proximity (and expression correlation) to be grouped together. This would create extra network nodes, potentially inflating the NDA scores in the permuted sets. To mitigate against this, those TSS from each permutation that do not conform to the input set distribution are re-entered into a further circular permutation until an identical distribution is found. If no matching grouping is found after 8 repeat permutations, additional regulatory regions are added from consecutive positions above and below whichever group is nearest in size to the relevant group in the original input dataset.
False discovery rates (FDR) are calculated using the Benjamini-Hochberg method.
The enrichment for GWAS hits from a pooled resource comprising the NCBI GWAS catalog and the GWASdb database (observed
NHGRI GWAS catalog, June 2014
GWASdb2, June 2014 update
Overlapping phenotypes, such as “urate” and “uric acid” were manually merged. Phenotypes that were considered to be too broad to be informative were excluded, as were those that were not related to human disease. A complete table of phenotypes in GWASdb and NCBI GWAS catalog, showing mergers and inclusion/exclusion in the present work, is provided in a supplementary file (
Strong anti-correlation between pairs of TSS associated with the same phenotype may have biological importance, such as down-regulation at one TSS but expression at another, or negative regulation of a signalling pathway on which expression of a TSS is dependent. For this reason, anti-correlations may improve detection of true associations in this analysis. However, in order to confer an overall improvement on the performance of the algorithm, true inverse expression relationships between phenotype-associated TSS would need to be sufficiently common to overcome the noise added by incorporating all strong anti-correlations into the NDA score. Anti-correlations do not contribute any net improvement to the NDA scores for a training set (Crohn’s disease, 50% of all SNPs, chosen at random), and were therefore excluded.
Full GWAS or meta-analysis data, reporting every SNP genotyped or imputed in a given study, are required in order to permute subsets against the appropriate background for a given study. These were obtained from the following sources:
Crohn’s disease summary
Ulcerative colitis summary
Summary
Summary
Summary
In order to better understand the pathophysiological implications of disease variants in regulatory regions, we sought to identify whether these regions exhibit unexpectedly specific expression in any given cell types or tissue samples. In order to reduce noise, technical and biological replicates were averaged for this and subsequent analyses. The full table of samples in FANTOM5, showing which samples were averaged as technical replicates, and which were excluded, is in
For a given trait, we took the subset of regulatory regions for which a significant coexpression pattern was detected for that trait (coexpression
There are several possible sources of bias in this raw measurement. For example, some cell types have more cell-type specific transcriptional activity, perhaps because these cell types fulfil a specialised role; other cell types are particularly well-represented in the FANTOM5 samples. We therefore controlled for the probability that a given cell type would be highly ranked in the initial RRA analysis, by permuting RRA results for at least 100,000 random selections of
Computer code required to run the NDA method, specifically for the detection of coexpression in FANTOM5 regulatory regions, can be obtained from
Our initial evaluation demonstrated that coexpression is stronger among regulatory regions containing variants with low GWAS p-values (
Coexpression signals are shown for six different –log10(p) bins for GWAS p values from a single study of Crohn’s disease. From each bin, 800 SNPs were selected at random. No signal for coexpression is detected at weak p-values.
The difference between the distributions of NDA scores derived from pre- and post-mapping permutations reveals the different components of the measure. When compared to a random pool of SNPs (pre-mapping permutations), two factors inflate the NDA scores for real GWAS data: firstly, more regulatory regions are identified because true GWAS hits are enriched within regulatory regions; secondly, the coexpression signal itself is greater for real data. In contrast, post-mapping permutations have precisely the same number of regulatory regions included as the real dataset, so there is no component of inflation due to enrichment in regulatory regions. The effects of these different components are shown in
Similar expression profiles are often seen arising from regulatory regions that are close to each other on the same chromosome, which may also span linkage disequilibrium blocks. The effect of this on the coexpression signal was mitigated by grouping nearby (within 100,000bp) regulatory regions into a single unit, unless they have notably different expression patterns. SNPs in nearby regulatory regions are also more likely to be in linkage disequilibrium, and these regulatory regions themselves are more likely to share cis- (or short-range trans-) regulatory signals in common. We checked for significant linkage disequilibrium between regulatory regions assigned to independent groups. At a threshold of r2 > 0.8, there is no linkage disequilibrium between significantly coexpressed groups; three examples of weaker linkage relationships were detected with 0.08 ≤ r2 ≤ 0.6 (Supplementary results).
Regulatory regions around individual TSS with higher coexpression scores contain variants with stronger GWAS p-values (
(a) Region surrounding IL10 (b) Region surrounding C1orf106. Top panel: Coloured rectangles show genomic location of individual regulatory regions (promoters or enhancers). Height of regulatory regions on y-axis depicts the coexpression score assigned to this regulatory region; groups of regulatory regions considered as a single unit (see
In order to enable the detection of new regulatory regions with strong coexpression relationships, we chose a permissive threshold at GWAS
Crohn's disease | 1924 (0.6) | 133 (3.5) | 5.7 | 70 | 23 (33%) | 1.61e-05 |
Ulcerative colitis | 2162 (0.7) | 146 (3.8) | 5.5 | 83 | 20 (24%) | 2.28e-06 |
LDL | 4644 (1.5) | 205 (5.2) | 3.5 | 92 | 19 (21%) | 1.48e-04 |
Total cholesterol | 6421 (2.0) | 316 (8.3) | 4.1 | 128 | 29 (23%) | 6.55e-07 |
Triglycerides | 4863 (1.5) | 254 (7.0) | 4.6 | 97 | 23 (24%) | 8.35e-06 |
Height | 8882 (2.8) | 358 (7.6) | 2.7 | 166 | 29 (17%) | 1.25e-06 |
HDL | 5410 (1.7) | 260 (7.2) | 4.2 | 101 | 17 (17%) | 3.51e-04 |
SBP | 417 (0.1) | 20 (0.4) | 3.0 | 13 | 0 (0%) | 4.89e-01 |
DBP | 711 (0.2) | 20 (0.4) | 1.9 | 14 | 0 (0%) | 5.41e-01 |
KS test: Kolmogorov-Smirnov test comparing distribution of coexpression scores for this study with permuted values.*Initial optimisation and parameterisation of the algorithm was undertaken using a random subset of data from this study.
Although many coexpressed regulatory regions are not promoters for annotated genes (supplementary results;
For a given disease, regulatory regions containing GWAS variants are coexpressed if they share similar activity patterns (i.e. similar expression patterns among transcripts arising from these regulatory regions) with other regulatory regions implicated in that disease.
(a) Region surrounding IL10 (b) Region surrounding C1orf106. Top panel: Coloured ectangles show genomic location of individual regulatory regions (promoters or enhancers). Height of regulatory regions on y-axis depicts the coexpression score assigned to this regulatory region; groups of regulatory regions considered as a single unit (see
We saw no evidence of spurious coexpression due to genomic proximity with shared regulatory influences (see
The coexpression signal essentially combines the signal for association in a GWAS with the location and activity pattern of regulatory regions on the genome. We deliberately chose a permissive GWAS p-value threshold in order to enable the detection of new signals that did not achieve genome-wide significance in the original studies. For example, we found that coexpressed transcripts for both LDL and total cholesterol (TC) arise from promoters for well-studied genes such as APOB[
The significantly-coexpressed networks detected here could be regarded as revealing the signature expression profile, at least within the FANTOM5 dataset, for a given disease or trait. We next explored whether these signature expression patterns reveal cell types or biological processes that may contribute to the trait or disease susceptibility.
We therefore ranked cell types and tissues by transcriptional activity for each of the significantly-coexpressed loci for each trait, and combined the rankings using a robust rank aggregation[
The development of high-throughput genotyping methods has led to an explosion of associations between genetic markers and human diseases[
We report relationships between numerous regulatory regions that are not associated with named genes–a restriction that has previously limited the transition from genetic discovery to biological understanding[
Even for those disease-associated variants that can be reliably assigned to a named gene, previous attempts to draw functional inferences have, by necessity, relied on published data[
Results for Crohn’s disease and ulcerative colitis were compared to the report by Huang
Secondly, our analysis extends current knowledge by revealing two distinct groups of significantly-coexpressed regulatory regions for each of these diseases, with differing expression profiles. For Crohn’s disease, one group is restricted to immune cells, particularly monocytes exposed to inflammatory stimuli, while another group of regulatory regions is active in epithelial cells. In contrast, cell type associations with ulcerative colitis were statistically significant in rectum, colon and intestine samples, and in a distinct group of immune cells: macrophages exposed to bacterial lipopolysaccharide (
In either case the predominance of each process in an individual patient is likely to have therapeutic relevance. For example, the highly variable clinical response to immunomodulatory therapies, such as methotrexate[
The data used for development and testing of the coexpression approach were from large meta-analyses that incorporate genotyping (or imputation) of genetic variants at extremely high resolution, increasing the probability that variants will be found within regulatory regions. In future, the availability of whole-genome sequencing can reasonably be expected to produce many additional high-quality datasets for coexpression analysis. In principle, the NDA approach can be generalised to any network in which it is desirable to quantify the proximity of a subset of nodes.
The scale, depth and breadth of the FANTOM5 expression atlas enable detection of subtle coexpression signals for regulatory regions that have previously been undetectable. The NDA approach developed here enables the identification of cell types and regulatory regions implicated in disease pathogenesis, and contributes a new independent signal to fine mapping of causative loci. As additional genetic studies become available at greater genotyping resolution, we anticipate that this method will detect new genetic associations with disease and coexpressed modules underlying pathogenesis. The NDA method will enable the identification of critical cell types and processes implicated in mechanisms of disease, and enable further genetic stratification of disease endotypes by underlying mechanism.
The FANTOM5 atlas is accessible from
An online service running the coexpression method is available at
Code delivering the NDA coexpression method is available at
(XLSX)
(TXT)
(XLSX)
(PDF)
(PDF)
We would like to express our gratitude for the diligence and professionalism of the entire FANTOM5 consortium and to the members of the IIBDGC group, GIANT consortium, and Global Lipids consortium for freely sharing their data. We are particularly grateful to the tens of thousands of patients and healthy volunteers who donated DNA and other material to these studies. We would like to thank GeNAS for data production.
The members of the FANTOM5 consortium are listed here: Alistair R. R. Forrest, Hideya Kawaji, Michael Rehli, J. Kenneth Baillie, Michiel J. L. de Hoon, Vanja Haberle, Timo Lassmann, Ivan V. Kulakovskiy, Marina Lizio, Masayoshi Itoh, Robin Andersson, Christopher J. Mungall, Terrence F. Meehan, Sebastian Schmeier, Nicolas Bertin, Mette Jørgensen, Emmanuel Dimont, Erik Arner, Christian Schmidl, Ulf Schaefer, Yulia A. Medvedeva, Charles Plessy, Morana Vitezic, Jessica Severin, Colin A. Semple, Yuri Ishizu, Robert S. Young, Margherita Francescatto, Intikhab Alam, Davide Albanese, Gabriel M. Altschuler, Takahiro Arakawa, John A. C. Archer, Peter Arner, Magda Babina, Sarah Rennie, Piotr J. Balwierz, Anthony G. Beckhouse, Swati Pradhan-Bhatt, Judith A. Blake, Antje Blumenthal, Beatrice Bodega, Alessandro Bonetti, James Briggs, Frank Brombacher, A. Maxwell Burroughs, Andrea Califano, Carlo V. Cannistraci, Daniel Carbajo, Yun Chen, Marco Chierici, Yari Ciani, Hans C. Clevers, Emiliano Dalla, Carrie A. Davis, Michael Detmar, Alexander D. Diehl, Taeko Dohi, Finn Drabløs, Albert S. B. Edge, Matthias Edinger, Karl Ekwall, Mitsuhiro Endoh, Hideki Enomoto, Michela Fagiolini, Lynsey Fairbairn, Hai Fang, Mary C. Farach-Carson, Geoffrey J. Faulkner, Alexander V. Favorov, Malcolm E. Fisher, Martin C. Frith, Rie Fujita, Shiro Fukuda, Cesare Furlanello, Masaaki Furuno, Jun-ichi Furusawa, Teunis B. Geijtenbeek, Andrew P. Gibson, Thomas Gingeras, Daniel Goldowitz, Julian Gough, Sven Guhl, Reto Guler, Stefano Gustincich, Thomas J. Ha, Masahide Hamaguchi, Mitsuko Hara, Matthias Harbers, Jayson Harshbarger, Akira Hasegawa, Yuki Hasegawa, Takehiro Hashimoto, Meenhard Herlyn, Kelly J. Hitchens, Shannan J. Ho Sui, Oliver M. Hofmann, Ilka Hoof, Fumi Hori, Lukasz Huminiecki, Kei Iida, Tomokatsu Ikawa, Boris R. Jankovic, Hui Jia, Anagha Joshi, Giuseppe Jurman, Bogumil Kaczkowski, Chieko Kai, Kaoru Kaida, Ai Kaiho, Kazuhiro Kajiyama, Mutsumi Kanamori-Katayama, Artem S. Kasianov, Takeya Kasukawa, Shintaro Katayama, Sachi Kato, Shuji Kawaguchi, Hiroshi Kawamoto, Yuki I. Kawamura, Tsugumi Kawashima, Judith S. Kempfle, Tony J. Kenna, Juha Kere, Levon M. Khachigian, Toshio Kitamura, S. Peter Klinken, Alan J. Knox, Miki Kojima, Soichi Kojima, Naoto Kondo, Haruhiko Koseki, Shigeo Koyasu, Sarah Krampitz, Atsutaka Kubosaki, Andrew T. Kwon, Jeroen F. J. Laros, Weonju Lee, Andreas Lennartsson, Kang Li, Berit Lilje, Leonard Lipovich, Alan Mackay-sim, Ri-ichiroh Manabe, Jessica C. Mar, Benoit Marchand, Anthony Mathelier, Niklas Mejhert, Alison Meynert, Yosuke Mizuno, David A. de Lima Morais, Hiromasa Morikawa, Mitsuru Morimoto, Kazuyo Moro, Efthymios Motakis, Hozumi Motohashi, Christine L.Mummery, Mitsuyoshi Murata, Sayaka Nagao-Sato, Yutaka Nakachi, FumioNakahara, Toshiyuki Nakamura, Yukio Nakamura, Kenichi Nakazato, Erik vanNimwegen, Noriko Ninomiya, Hiromi Nishiyori, Shohei Noma, TadasukeNozaki, Soichi Ogishima, Naganari Ohkura, Hiroko Ohmiya, Hiroshi Ohno, Mitsuhiro Ohshima, Mariko Okada-Hatakeyama, Yasushi Okazaki, Valerio Orlando, Dmitry A. Ovchinnikov, Arnab Pain, Robert Passier, Margaret Patrikakis, Helena Persson, Silvano Piazza, James G. D.Prendergast, Owen J. L. Rackham, Jordan A. Ramilowski, Mamoon Rashid, Timothy Ravasi, Patrizia Rizzu, Marco Roncador, Sugata Roy, Morten B.Rye, Eri Saijyo, Antti Sajantila, Akiko Saka, Shimon Sakaguchi, MizuhoSakai, Hiroki Sato, Hironori Satoh, Suzana Savvi, Alka Saxena, Claudio Schneider, Erik A. Schultes, Gundula G. Schulze-Tanzil, Anita Schwegmann, Thierry Sengstag, Guojun Sheng, Hisashi Shimoji, Yishai Shimoni, Jay W. Shin, Christophe Simon, Daisuke Sugiyama, Takaaki Sugiyama, Masanori Suzuki, Naoko Suzuki, Rolf K. Swoboda, Peter A. C.’t Hoen, Michihira Tagami, Naoko Takahashi, Jun Takai, Hiroshi Tanaka, Hideki Tatsukawa, Zuotian Tatum, Mark Thompson, Hiroo Toyoda, TetsuroToyoda, Eivind Valen, Marc van de Wetering, Linda M. van den Berg, Roberto Verardo, Dipti Vijayan, Ilya E. Vorontsov, Wyeth W. Wasserman, Shoko Watanabe, Christine A. Wells, Louise N. Winteringham, Ernst Wolvetang, Emily J. Wood, Yoko Yamaguchi, Masayuki Yamamoto, Misako Yoneda, Yohei Yonekura, Shigehiro Yoshida, Susan E. Zabierowski, Peter G. Zhang, XiaobeiZhao, Silvia Zucchelli, Kim M. Summers, Harukazu Suzuki, Carsten O. Daub, Jun Kawai, Peter Heutink, Winston Hide, Tom C. Freeman, Boris Lenhard, Vladimir B. Bajic, Martin S. Taylor, Vsevolod J. Makeev, Albin Sandelin, David A. Hume, Piero Carninci, Yoshihide Hayashizaki.
The members of the International IBD Genetics Consortium are listed here: Murray Barclay, Laurent Peyrin-Biroulet, Mathias Chamaillard, Jean-Frederick Colombel, Mario Cottone, Anthony Croft, Renata D'Incà, Jonas Halfvarson, Katherine Hanigan, Paul Henderson, Jean-Pierre Hugot, Amir Karban, Nicholas A Kennedy, Mohammed Azam Khan, Marc Lémann, Arie Levine, Dunecan Massey, Monica Milla, Grant W Montgomery, Sok Meng Evelyn Ng, Ioannis Oikonomou, Harald Peeters, Deborah D. Proctor, Jean-Francois Rahier, Rebecca Roberts, Paul Rutgeerts, Frank Seibold, Laura Stronati, Kirstin MTaylor, Leif Törkvist, Kullak Ublick, Johan Van Limbergen, Andre VanGossum, Morten H. Vatn, Hu Zhang, Wei Zhang, Australia and New Zealand IBDGCJane M. Andrews, Peter A. Bampton, Murray Barclay, Timothy H. Florin, RichardGearry, Krupa Krishnaprasad, Ian C. Lawrance, Gillian Mahy, Grant W.Montgomery, Graham Radford-Smith, Rebecca L. Roberts, Lisa A. SimmsLeila Amininijad, Isabelle Cleynen, Olivier Dewit, Denis Franchimont, MichelGeorges, Debby Laukens, Harald Peeters, Jean-Francois Rahier, Paul Rutgeerts, Emilie Theatre, André Van Gossum, Severine Vermeire.Guy Aumais, Leonard Baidoo, Arthur M. Barrie III, Karen Beck, Edmond-JeanBernard, David G. Binion, Alain Bitton, Steve R. Brant, Judy H. Cho, AlbertCohen, Kenneth Croitoru, Mark J. Daly, Lisa W. Datta, Colette Deslandres, Richard H. Duerr, Debra Dutridge, John Ferguson, Joann Fultz, PhilippeGoyette, Gordon R. Greenberg, Talin Haritunians, Gilles Jobin, SeymourKatz, Raymond G. Lahaie, Dermot P. McGovern, Linda Nelson, Sok MengNg, Kaida Ning, Ioannis Oikonomou, Pierre Paré, Deborah D. Proctor, MiguelD. Regueiro, John D. Rioux, Elizabeth Ruggiero, L. Philip Schumm, MarcSchwartz, Regan Scott, Yashoda Sharma, Mark S. Silverberg, Denise Spears, A.Hillary Steinhart, Joanne M. Stempak, Jason M. Swoger, Constantina Tsagarelis, Wei Zhang, Clarence Zhang, Hongyu Zhao.Jan Aerts, Tariq Ahmad, Hazel Arbury, Anthony Attwood, Adam Auton, Stephen G Ball, Anthony J Balmforth, Chris Barnes, Jeffrey C Barrett, InêsBarroso, Anne Barton, Amanda J Bennett, Sanjeev Bhaskar, Katarzyna Blaszczyk, John Bowes, Oliver J Brand, Peter S Braund, Francesca Bredin, GeromeBreen, Morris J Brown, Ian N Bruce, Jaswinder Bull, Oliver S Burren, JohnBurton, Jake Byrnes, Sian Caesar, Niall Cardin, Chris M Clee, Alison J Coffey, John MC Connell, Donald F Conrad, Jason D Cooper, Anna F Dominiczak, Kate Downes, Hazel E Drummond, Darshna Dudakia, Andrew Dunham, Bernadette Ebbs, Diana Eccles, Sarah Edkins, Cathryn Edwards, Anna Elliot, Paul Emery, David M Evans, Gareth Evans, Steve Eyre, Anne Farmer, I Nicol Ferrier, Edward Flynn, Alistair Forbes, Liz Forty, Jayne A Franklyn, Timothy M Frayling, Rachel M Freathy, Eleni Giannoulatou, Polly Gibbs, Paul Gilbert, Katherine Gordon-Smith, Emma Gray, Elaine Green, Chris J Groves, Detelina Grozeva, Rhian Gwilliam, Anita Hall, Naomi Hammond, Matt Hardy, Pile Harrison, Neelam Hassanali, Husam Hebaishi, Sarah Hines, Anne Hinks, Graham A Hitman, Lynne Hocking, Chris Holmes, Eleanor Howard, PhilipHoward, Joanna MM Howson, Debbie Hughes, Sarah Hunt, John D Isaacs, Mahim Jain, Derek P Jewell, Toby Johnson, Jennifer D Jolley, Ian R Jones, Lisa A Jones, George Kirov, Cordelia F Langford, Hana Lango-Allen, G MarkLathrop, James Lee, Kate L Lee, Charlie Lees, Kevin Lewis, Cecilia MLindgren, Meeta Maisuria-Armer, Julian Maller, John Mansfield, Jonathan LMarchini, Paul Martin, Dunecan CO Massey, Wendy L McArdle, PeterMcGuffin, Kirsten E McLay, Gil McVean, Alex Mentzer, Michael LMimmack, Ann E Morgan, Andrew P Morris, Craig Mowat, Patricia BMunroe, Simon Myers, William Newman, Elaine R Nimmo, Michael CO'Donovan, Abiodun Onipinla, Nigel R Ovington, Michael J Owen, KimmoPalin, Aarno Palotie, Kirstie Parnell, Richard Pearson, David Pernet, John RB Perry, Anne Phillips, Vincent Plagnol, Natalie J Prescott, Inga Prokopenko, Michael A Quail, Suzanne Rafelt, Nigel W Rayner, David M Reid, AnthonyRenwick, Susan M Ring, Neil Robertson, Samuel Robson, Ellie Russell, David St Clair, Jennifer G Sambrook, Jeremy D Sanderson, Stephen J Sawcer, Helen Schuilenburg, Carol E Scott, Richard Scott, Sheila Seal, Sue ShawHawkins, BeverleyM Shields, Matthew J Simmonds, Debbie J Smyth, ElilanSomaskantharajah, Katarina Spanova, Sophia Steer, Jonathan Stephens, Helen EStevens, Kathy Stirrups, Millicent A Stone, David P Strachan, Zhan Su, Deborah PM Symmons, John R Thompson, Wendy Thomson, Martin D Tobin, Mary E Travers, Clare Turnbull, Damjan Vukcevic, Louise V Wain, MarkWalker, Neil M Walker, Chris Wallace, Margaret Warren-Perry, Nicholas AWatkins, John Webster, Michael N Weedon, Anthony G Wilson, MatthewWoodburn, B Paul Wordsworth, Chris Yau, Allan H Young, EleftheriaZeggini, Matthew A Brown, Paul R Burton, Mark J Caulfield, AlastairCompston, Martin Farrall, Stephen CL Gough, Alistair S Hall, Andrew THattersley, Adrian VS Hill, Christopher G Mathew, Marcus Pembrey, JackSatsangi, Michael R Stratton, Jane Worthington, Matthew E Hurles, Audrey Duncanson, Willem H Ouwehand, Miles Parkes, Nazneen Rahman, John ATodd, Nilesh J Samani, Dominic P Kwiatkowski, Mark I McCarthy, Nick Craddock, Panos Deloukas, Peter Donnelly, Jenefer M Blackwell, Elvira Bramon, Juan P Casas, Aiden Corvin, Janusz Jankowski, Hugh SMarkus, Colin NA Palmer, Robert Plomin, Anna Rautanen, Richard CTrembath, Ananth C Viswanathan, Nicholas W Wood, Chris C A Spencer, Gavin Band, Céline Bellenguez, Colin Freeman, Garrett Hellenthal, EleniGiannoulatou, Matti Pirinen, Richard Pearson, Amy Strange, HannahBlackburn, Suzannah J Bumpstead, Serge Dronov, Matthew Gillman, Alagurevathi Jayakumar, Owen T McCann, Jennifer Liddle, Simon C Potter, Radhi Ravindrarajah, Michelle Ricketts, Matthew Waller, Paul Weston, SaraWidaa, Pamela Whittaker.