ICMS

May 7, 2007 - May 11, 2007

ICMS, 14 India Street, Edinburgh

Organisers

Name	Institution
Baur, Max	University of Bonn
Sheehan, Nuala	University of Leicester
Thompson, Elizabeth A	University of Washington

Short Report

Genetic Epidemiology is such an active area of research at the moment that there is a serious shortage of expertise in the UK to meet the current demands. Highly sophisticated statistical methods are required to deal with the complexity that arises in many biological applications due to large datasets, indirect measurements, complex underlying biological processes and various combinations of all three. Moreover, recent developments in large scale bioscience are rapidly out-pacing current analytic capacity. Strong national and international collaborations are crucial if UK scientists are to stay up to date in this rapidly developing area. The aim of this workshop was to gather together a sufficiently expert group with interests ranging across diverse areas of genetic epidemiological application in order to focus on the computational and methodological issues that are common to family studies, genetic association studies and the analyses of complex traits. Participants thus had expertise in Markov chain Monte Carlo methods, graphical modelling, statistical computation and in the management and analysis of large applied population studies. One objective was to assess existing methods of analysis of large and complex datasets, to consider appropriate adaptation of these methods and to investigate the potential for new methods which address common problematic features of all these datasets. Another was to actively encourage young researchers into this field by providing the opportunity to become acquainted with the main research issues in the area and to make vital contacts with the leading people and their groups in an informal small group setting.

Participants list and links to available presentations are further down this page.

Download the pdf file of the full report

Arrangements

Participation
Participation is by invitation only and, due to the limited number of places available, it is desirable that participants attend the full workshop. The workshop will begin with lunch on Monday 7 May and finish after lunch on Friday 11 May 2007.

UK Visas
If you are travelling from overseas you may require an entry visa. A European visa does not guarantee entry to the UK. Please use this link to the UK Visas site to find out if you need a visa and if so how to apply for one. If you do require a visa, ICMS can provide a signed invitation letter.

Venue
The workshop will take place at the head-quarters of ICMS, 14 India Street, Edinburgh. This house is the birthplace of James Clerk Maxwell and is situated in the historic New Town of Edinburgh, near the city centre.

The ICMS travel pages contain advice on how to travel to Edinburgh. For local information the finding ICMS page shows the location of ICMS and contains useful maps of the city centre.

The seminar room at ICMS has whiteboards, 2 overhead projectors, a data projector and laptop.

Wireless access is available throughout the ICMS building. There are also 7 public PCs which may be used at any time for internet access and to check email.

Accommodation
ICMS will arrange rooms in local guest houses for those who require it. Accommodation is typically about 15 to 30 minutes walk from ICMS. Participants are also free to make their own arrangements and may claim back the cost, with receipts, up to a maximum of £45.00 per night bed and breakfast. A list of Edinburgh accommodation of various sorts and prices is available here. Sections 1-3 are particularly relevant.

Meals and Refreshments
The workshop will open with a sandwich lunch between 12.00 and 13.30 on Monday 7 May and close after a light lunch on Friday 11 May. For the remainder of the days, participants are free to go out for lunch and explore the many cafes, restaurants, sandwich shops and bars in the surrounding area.

Morning and afternoon refreshments will be provided throughout the workshop.

There will be an informal wine reception after the close of lectures on Monday 7 May evening.

Evening meals will be provided in local restaurants on Tuesday 8 and Wednesday 9 May. The workshop dinner will take place on the evening of Thursday 10 May. No evening meal has been included on the first day as participants may wish an early night after travelling. On arrival you will be given a ‘welcome pack’ which contains a list of local restaurants.

Registration
Registration will take place between 12.00 and 13.30 on Monday 7 May, with a sandwich lunch provided. The talks will start at 13.30.

Posters
Post-doctoral researchers wishing to display posters will be able to do so in the Exhibition Room at ICMS from Tuesday afternoon until Friday morning. Two poster sessions have been programmed - Wednesday 9 May and Thursday 10 May from 17.30 until 18.30 - for any participants wishing to view the posters and discuss the research.

Financial Arrangements
Unless otherwise specified in your invitation letter, the workshop grant will cover the cost of 4 nights’ bed and breakfast accommodation, lunch on 2 days, evening meals on 3 nights and the informal wine reception (see ‘Meals and Refreshments’ section above).

We understand that some participants may be able to cover some or all of their own travel. For others, we would hope to be able to reimburse ecomony/second class travel to Edinburgh. Details will be in your individual invitation letter and information may be updated once participants have completed the ‘travel estimates’ section of the on-line registration form. Reimbursement will take place after the meeting and you will be paid directly into your bank account in the currency of your choice. At the workshop you will be given a claim form. Please note that receipts are required for every item claimed (we can make copies of your travel tickets).

Under the terms of our EPSRC funding we are required to charge a 30.00 GBP registration fee to cover costs not admissible under the grant. The fee will be payable on arrival at the workshop. Payment may be by cash, sterling cheque or credit/debit card (it may not be claimed back from ICMS). If you anticipate any difficulty covering the fee, please let me know.

Programme

Monday 7 May

12.00 – 13.30	Buffet lunch and Registration
13.30 – 15.30	Overview Talks David Clayton (University of Cambridge) Genome-wide association studies; experiences of the WTCCC Duncan Thomas (University of Southern California, Los Angeles) Complex biological pathways
15.30 – 16.00	Coffee/tea
16.00 – 18.00	Multifactorial traits Elizabeth A Thompson (University of Washington) Assessing the significance of linkage signals Download presentation Laura Almasy (Southwest Foundation for Biomedical Research (SFBR)) Quantitative risk factors for identifying genes in complex diseases Francoise Clerget Darpoux (INSERM & Paris-Sud University) Multifactorial diseases: a gap between association information and the understanding of the pathogenic process Download presentation
18.30 – 20.00	Wine Reception at ICMS, 14 India Street (No group dinner arranged for this evening)

Tuesday 8 May

9.00 – 10.30	Graphical Models & Causality Vanessa Didelez (University College London) Causal inference for genetic epidemiology using Mendelian randomisation Download presentation Alun Thomas (University of Utah) Towards linkage analysis with markers in linkage disequilibrium by graphical modelling Download presentation
10.30 – 11.00	Coffee/tea
11.00 – 12.30	Relationship Estimation Elja Arjas (University of Helsinki) Estimating genealogies from marker data: a Bayesian approach Simon Heath (National Genotyping Centre, France) Detection/reconstruction of family relationships using genotype data from whole genome arrays
12.30 – 14.00	Lunch
14.00 – 15.30	Discussion Session Pedigrees and Family Data Panellists: Elizabeth Thompson, Cornelia Van Duijn, Simon Heath
15.30 – 16.00	Coffee/tea
16.00 – 18.00	Linkage Analysis Cornelia Van Duijn (Erasmus University Medical Centre) Splitting complex pedigrees for linkage analysis Download presentation (6MB) Thore Egeland (Ullevål University Hospital) Adjusting for relatedness among founders in linkage analysis Download presentation Konstantin Strauch (Philipps University Marburg) A close view of the possible triangle Download presentation
19.00	Dinner at Nargile Restuarant, 73 Hanover Street

Wednesday 9 May

9.00 – 10.30	Quantitative traits John Whittaker(London School of Hygiene and Tropical Medicine) Baysian meta-analysis of genetic association studies Heather Cordell (Newcastle University) Testing and estimation of genotype and haplotype effects in family-based analysis of quantitative traits with missing genotype data
10.30 – 11.00	Coffee/tea
11.00 – 12.30	Study design and analysis Juni Palmgren (Karolinska Institutet, Stockholm) The GeneStat web portal
12.30 – 14.00	Lunch
14.00 – 15.30	Discussion Session Genome Wide Association Panellists: Duncan Thomas, David Clayton, Peter Holmans, Heike Bickeböller
15.30 – 16.00	Coffee/tea
16.00 – 17.30	Small Group Sessions (to be arranged)
17.30 – 18.30	Poster session
19.00	Dinner at Balli's Bistro, 89 Hanover Street

Thursday 10 May

9.00 – 10.30	DNA Structure/Function Wally Gilks (University of Leeds) DNA is not a straight line Download presentation (4MB) Dawn Teare(University of Sheffield) Study of a candidate copy number polymorphism in asthma families
10.30 – 11.00	Coffee/tea
11.00 – 12.30	Small Group Sessions (to be arranged)
12.30 – 14.00	Lunch
14.00 – 15.30	Discussion Session Data integration Panellists: Robert Elston, John Whittaker, Wally Gilks
15.30 – 16.00	Coffee/tea
16.00 – 17.30	Applications Sarah Lewis (University of Bristol) Some practical examples illustrating the use of Mendelian randomisation Tim Bishop (Cancer Research UK, University of Leeds) Challenges to understanding disease aetiology with family studies
17.30 – 18.30	Poster Session
19.00	Workshop Dinner at First Coast Restaurant, 99-101 Dalry Road (dress casual)

Friday 11 May

9.00 – 10.30	Covariates and Population Stratification Peter Holmans (University of Wales College of Medicine, Cardiff) Use of covariates in model-free linkage and association Download presentation David Balding(Imperial College London) Population structure and genetic associations
10.30 – 11.00	Coffee/tea
11.00 – 12.30	Grand Wrap-Up Robert Elston (Case Western Reserve University) Summary and Conclusions Download presentation
12.30	Buffet Lunch and End of Workshop

Presentations:

Presentation Details
Almasy, Laura
Quantitative risk factors for identifying genes in complex diseases
View Abstract Hide Abstract
One common approach to localizing and identifying genes influencing common complex diseases is to study quantitative risk factors related to the disease. Such studies have been relatively successful in localizing human quantitative trait loci (QTLs) and, recently, in identifying potentially functional quantitative trait nucleotides (QTNs). In this talk, I will review the rationale behind using quantitative risk factors for genetic studies of complex disease and will discuss variance component methods for linkage and association studies of quantitative traits, with examples from ongoing studies.
Arjas, Elja
Estimating genealogies from marker data: a Bayesian approach
View Abstract Hide Abstract
An issue often encountered in statistical genetics is whether, or to what extent, it is possible to estimate the degree in which individuals sampled from a background population are related to each other, on the basis of the available diploid multi-locus genotype data and some information on the demography of that population. In this talk, this question is considered by using an explicit modelling and reconstruction of the pedigrees and gene flows at the marker loci. For computational reasons, the analysis is restricted to a relatively recent history of the population, currently extending, depending of the data, up to ten or twenty generations backwards in time. As a computational tool, we use Markov Chain Monte Carlo numerical integration on the state space of genealogies of the sample individuals. The main technical challenge has been in devising a variety of joint proposal distributions which would guarantee that the algorithm has reasonable mixing properties. As illustrations of the method, we consider the question of relatedness both in terms of individuals (pedigree based relatedness estimation) and at the level of genes/genomes (IBD-estimation), using both simulated and real data. (Joint work with Dario Gasbarra, Matti Pirinen and Mikko Sillanpää)
Balding, David
Population structure and genetic associations
View Abstract Hide Abstract
I will review the rapidly-developing field of statistical methods to account for the effects of population structure in genetic association studies, and present a new method with good properties. I argue that with many genome-wide markers, the problem can be more-or-less fully resolved.
Bishop, Tim
Challenges to understanding disease aetiology with family studies
View Abstract Hide Abstract
While approaches to understanding disease aetiology when examining a single phenotype are well developed, examination of multiple phenotypes should help sort out the interrelationship between the phenotypes and the effect of genes and environment are less well developed. The complexities are illustrated with examples from the studies of cancer and associated phenotypes.
Clayton, David
Genome-wide association studies; experiences of the WTCCC
View Abstract Hide Abstract
This talk will be an overview of the problems and achievements of the Wellcome Trust Case-Control Consortium study, the main part of which involves 7 disease groups, 2 control groups and 17,000 subjects typed for 500,000 SNPs.
Clerget-Darpoux, Françoise
Multifactorial diseases: a gap between association information and the understanding of the pathogenic process
View Abstract Hide Abstract
Genome-wide association tests using tagSNPs are becoming an all-purpose tool for the study of common, multifactorial diseases, despite the poor information provided on the effects of causal variants by the associated markers. We show here that, even in the ideal situation where a single causal variant underlies an association between a disease and a tagSNP, the available information is too poor to infer the genotype relative risks (GRR) of the causal variant. Additionally, under more realistic scenarios in which many variants interact together, there is a long way to go between establishing an allelic (or haplotypic) association and being able to model the effects of the causal variants. This is not to say that genome-wide association studies cannot bring significant contributions to the understanding of diseases. There are, however, reasons to think that the expectations they have raised are inflated and, most importantly, may lead to the neglect of other important and complementary strategies.
Cordell, Heather
Testing and estimation of genotype and haplotype effects in family-based analysis of quantitative traits with missing genotype data
View Abstract Hide Abstract
A convenient method for testing genetic association with a quantitative trait is to perform a linear regression of trait on variables coding for the genotype or haplotypes present in an individual. Such analysis is not generally valid in the presence of population stratification, but a valid analysis can be achieved with family data by incorporation of additional stratification parameters based on the mating type of an individual's parents. Correlations between individuals from the same family (due to additional unmeasured genetic or shared environmental effects) can be accounted for by use of an empirical variance estimate clustered by family. In practice, many individuals may have missing genotype data and/or unobserved phase (haplotype) resolutions, although the possible configurations may be inferred probabilistically given genotype data from other family members and/or by taking account of linkage disequilibrium with neighboring loci. A valid analysis can be achieved by use of a conditioning argument that essentially discards individuals or families in which the required information is not inferable, but at the cost of some loss in power and efficiency. Instead, we propose incorporation of the posterior probabilities of the possible configurations through a weighted regression procedure.
Didelez, Vanessa
Causal inference for genetic epidemiology using Mendelian randomisation
View Abstract Hide Abstract
Mendelian Randomisation means the random assignment of genotype from parent to offspring. When the causal effect of a modifiable phenotype on a disease outcome is known to be confounded by unobservable quantities, the corresponding genotype can sometimes be used as an instrumental variable (IV) to estimate this causal effect because we may reasonably assume - due to the random assignment - that genotype and those unobservable quantities are independent. In this talk I will give an introduction to causal inference with instrumental variables, and show how graphical models can be used to illustrate and verify the required assumptions for typical Mendelian randomisation applications. Standard IV methods are often based on linear models which are usually not appropriate for binary disease outcomes and hence some alternative models and methods of estimation will be discussed.
Egeland, Thore
Adjusting for relatedness among founders in linkage analysis
View Abstract Hide Abstract
The aim of linkage analysis is to find genes responsible for diseases based on family materials. In many applications there is uncertainty about the exact pedigree. For instance, parents may or may not be first-cousins. The results of the analyses depend on the specification of the pedigree. Failing to specify the correct family structure may inflate lod scores in cases where there is no linkage and reduce the strength of signal when there is linkage (Miano et al., 2000; Hossjer, 2006). The approach of this talk is to list possible pedigrees, specify prior probabilities, calculate posterior probabilities as in Sheehan and Egeland (2007) and adjust the lod-scores according to the posteriors. The method is exemplified using simulated data analysed by available freeware. (joint work with Nuala Sheehan, University of Leicester) References Hossjer, O. Modeling the effect of inbreeding among founders in linkage analysis. Theor Popul Biol. 2006 Sep;70(2):146-63; Miano MG, Jacobson SG, Carothers A, Hanson I,Teague P, Lovell J, Cideciyan AV, Haider N, Stone EM, Sheffield VC, Wright AF. Am J Hum Genet. 2000 Nov;67(5):1348-51; Sheehan, NA and T. Egeland. Structured incorporation of prior information in relationship identification problems. Ann Hum Genet. 2007 Jan 17 [Epub ahead of print]
Gilks, Wally
DNA is not a straight line
View Abstract Hide Abstract
We often think of the genome as a long string of nucleotides, with SNPs scattered along it. However, genomic DNA is a molecule which lives in 4-dimensional space-time, just like any other molecule. It organises itself structurally to form long-range interactions with itself which are vital to the functioning of the cell and the organism. Some of these interactions are dynamic and some are conserved through evolution. Ultimately, the field of statistical genetics will have to take all this into account when building models of the cause of disease. In this non-statistical presentation, I will review some emerging structural and dynamic models of DNA based on recent experimental and bioinformatic evidence.
Heath, Simon
Detection/reconstruction of family relationships using genotype data from whole genome arrays
View Abstract Hide Abstract
Using the large amounts of information from whole genome arrays, close relationships such as parent-offspring and sibs can be easily detected and distinguished from each other. As relationships become more distant, however, the differences between different relationship types become more subtle. I will discuss to what extent more distant relationships can be distinguished using the current generation of mapping arrays.
Holmans, Peter
Use of covariates in model-free linkage and association
View Abstract Hide Abstract
Complex traits often exhibit phenotypic heterogeneity (for example, positive and negative symptoms in schizophrenia). The power of linkage and association analyses of complex traits will be improved if this heterogeneity is taken into account. A promising way to do this is to include the various phenotypes in the analysis as covariates. This talk will give a review of commonly-used methods, together with situations where these are particularly advantageous. For association methods, the situation where covariates are only measurable in the cases is of particular interest.
Lewis, Sarah
Some practical examples illustrating the use of Mendelian randomisation
View Abstract Hide Abstract
Observational epidemiology is plagued by problems of confounding, bias, measurement error and reverse causation, associations arising can therefore often be misleading. However, common polymorphisms exist which influence; exposure propensities, metabolism of environmental exposures or cellular response to specific exposures. The distribution of alleles, and consequently genotypes is largely unrelated to the sorts of confounders and biases that may distort interpretations of findings from observational studies. In addition genotypes are assigned at birth and associations between genotypes and disease are not subject to reverse causation. Therefore, determination of genotype can be used and related to disease risk in some incidences as a surrogate for directly measuring the exposure of interest. This approach is of value when suitable polymorphisms exist which modify the biologically active dose of the exposure of interest, and provides new opportunities to test causality and explore biological mechanisms. This is an illustration of how investment in molecular epidemiology may contribute to understanding and preventing the adverse effects on human health of modifiable exposures. Some examples of this approach termed "Mendelian randomization" will be presented, including associations between the ALDH2 gene and blood pressure, in addition I will discuss of the strength and limitations of this tool.
Palmgren, Juni
The GeneStat web portal
View Abstract Hide Abstract
I will give a brief presentation of the rationale and current status of the GeneStat web portal as a tool for disseminating information on the design and analysis of genetic studies. GeneStat was created in 2004 to help clinicians, geneticists and statisticians orientate in the myriad of new methods and genetic software made available to the end user. The early focus was on genetic association studies. In 2007 the development of GeneStat has been taken up in the context of European Biobanking. The site (http://dev.samuliripatti.com/genestat/) is under revision and discussions are ongoing concerning the need for, the format of and the ultimate home for such an information portal. The aim of this talk is to bring this discussion to the current meeting.
Strauch, Konstantin
A close view of the possible triangle
View Abstract Hide Abstract
In a MOD-score analysis, the parametric LOD score is maximized with respect to the trait-model parameters.  For affected sib pairs, the MOD score is identical to the possible-triangle test statistic, and so it asymptotically follows a mixture of chi-square distributions with 2, 1 and 0 degrees of freedom under the null hypothesis of no linkage.  In this context, some (albeit limited) information regarding the trait-model parameters can be gained, and there is a chance for an increased power to detect linkage compared to a simple LOD-score analysis.  It can be shown more generally that, for any given type of pedigree, the parametric MOD score and the corresponding nonparametric likelihood-based linkage test statistic are the same.  The number of degrees of freedom underlying the null distribution depends on the type of pedigrees studied.  This number corresponds to the number of explored dimensions related to power and to the number of parameters that can jointly be estimated.
Teare, Dawn
Study of a candidate copy number polymorphism in asthma families.
View Abstract Hide Abstract
Understanding the genetic basis of disease phenotype variation is a major goal in genetic epidemiology. Genome-wide association studies provide a powerful framework to screen for risk modifying genetic variants, allowing candidate loci arising from such screens to be considered in a family based setting for gene characterisation and models of gene/gene and gene/environment interaction. While segregation or linkage methods considering allelic states are well established, these methods generally assume codominant markers and diploid loci and are not well-suited to studying other forms of genetic variation such as structural variants and allelic expression phenotypes. Copy number variants (CNVs) are emerging as an important component of genetic variation and characteristics of CNVs are currently being assessed through genome wide SNP technology platforms. We will present an overview of the features and forms of genetic variation that may need to be considered within a candidate locus (or genomic region) family based analysis, and the demands this will place on the statistical model and analysis.
Thomas, Alun
Towards linkage analysis with markers in linkage disequilibrium by graphical modelling
View Abstract Hide Abstract
We review recent developments of MCMC integration methods for computations on graphical models for two applications in statistical genetics:modelling allelic association and pedigree based linkage analysis. We discuss and illustrate estimation of graphical models from haploid and diploid genotypes, and the importance of MCMC updating schemes beyond what is strictly necessary for irreducibility. We then outline an approach combining these methods to compute linkage statistics when alleles at the marker loci are in linkage disequilibrium. Other extensions suitable for analysis of SNP genotype data in pedigrees are also discussed and programs that implement these methods, and which are available from the author's web site, are described. We conclude with a discussion of how this still experimental approach might be further developed.
Thomas, Duncan
Complex biological pathways
View Abstract Hide Abstract
While it is widely understood the "complex diseases" result from the interplay of multiple genes and environmental agents, there is little consensus about how to go about addressing this problem analytically. In this talk, I will provide an overview of a range of options, both exploratory and model-based. In the context of a genome-wide association study, for example, one is unlikely to have any strong biological priors for most of the hundreds of thousands of markers under study, so the search for main effects and interactions is largely an exploratory exercise. That said, it may be possible to incorporate prior knowledge into a hierarchical modeling framework to help prioritize the selection of SNPs from the first stage for testing in later stages or to select the final list of associations to report at the end. Likewise, one must decide whether to search for all possible pairwise or higher-order gene-gene interactions or target the search in some way. These issue will be discussed in more detail in the Wednesday afternoon panel discussion. In the Monday afternoon talk, I plan to focus on a more limited set of genes and environmental exposures that comprise one or more candidate pathways based on a priori biological knowledge (or hypotheses). Despite such a focused subset of variables, one could still proceed in an exploratory mode (e.g., using CART, MDR, Neural Nets, Random Forrets, or a host of other tools). However, if one begins with a biologically-related hypothesis, it makes sense that one might want to incorporate this into the modeling itself in some manner. I will discuss two main approaches to accomplishing this, one based on mechanistic models (e.g., differential equations models based on Michalis-Menten kinetics for metabolic pathways), one based more empirically on hierarchical Bayes models in which the biological knowledge enters as prior covariates in the higher levels of the model. The latter allows one to take account of uncertainty about the terms to include in the model in a natural way through a form of Bayesian model averaging. One might also marry the two approaches by using the mechanistic model predictions as prior covariates in a hierarchical model. Finally, I will discuss a general framework in terms of complex models involving a network of latent variables where the topology of that network is itself uncertain, but again informed by biological knowledge, organized in a formal ontology.
Thompson, Elizabeth A
Assessing the significance of linkage signals
View Abstract Hide Abstract
In 1955, it was suggested that a base-10 lod score of 3 was sufficient evidence of linkage, and over the next 40 years that criterion lead to the construction of genome-wide human genetic maps and the mapping of many genetically simple traits. However, in the era of complex traits and the multiple testing of genome scans, we have no good measure of the significance of a maximum lod score and null distributions of many linkage detection test statistics are unobtainable. Asymptotic results are often cited, but the applicability of these is likewise unclear. Here we present a new approach that can not only provide valid p-values, but also have some robustness to marker model uncertainty, and further provides a measure of uncertainty as to the strength of the evidence resulting from uncertainty in the underlying inheritance at marker loci, given marker data.
van Duijn, Cornelia M
Splitting complex pedigrees for linkage analysis
View Abstract Hide Abstract
Linkage analysis is expected to be a powerful tool for the localization and isolation of genes. Although families have become scarce, genetically isolated populations offer still opportunities for linage studies. Also for complex traits, large pedigrees can be constructed often spanning 10-15 generations. However, computational problems occur when conducting linkage studies of such complex pedigrees, since patients are often related through multiple lines and phenotypic information is lacking for most ancestors. We have developed a new method to cut large pedigrees into sub-pedigrees. We developed a recursive algorithm based on 3 criteria: (1) within a sub-pedigree the kinship is maximized, (2) a patient can be assigned to only one sub-pedigree, (3) the size of pedigree can be restricted by a user specified bit size. We applied our algorithm to a pedigree including 103 Alzheimer patients from the Genetic Research in Isolated Populations (GRIP) program, which is based in the Southwest of the Netherlands. The pedigree spanning 17 generations was cut into 35 sub-pedigrees with on average 3 patients. We conducted 100 genome-wide simulations under no linkage, which showed that a LOD score of 3.63 in these sub-pedigrees corresponded to a genome wide type 1 error of 0.05. Two regions passed this threshold in the AD genome screen using 420 micro-satellite markers, one known region on chromosome 1 and one new region on chromosome. Both regions were confirmed by dense fine typing suggesting we have developed a powerful algorithm for splitting complex extended pedigrees. (Joint work with Fan Liu, Yurii S Aulchenko, Erasmus University Medical School)
Whittaker, John
Bayesian meta-analysis of genetic association studies
View Abstract Hide Abstract
Reliable assessment of genetic effects on quantitative traits or complex disease risk often requires synthesis of evidence from all available association studies to achieve adequate sample size. Frequently, individual studies will have genotyped single or partially overlapping sets of SNPs within a gene or region of interest, hampering attempts to combine all the available data. Using the example of C-reactive protein (CRP) as a quantitative trait, we show how information on linkage disequilibrium in its gene facilitates use of Bayesian hierarchical models to integrate informative data from all available genetic association studies of this trait, irrespective of the SNP typed. Bayesian variable selection, followed by contextualisation of SNPs exhibiting independent associations within the haplotype structure of the gene enhanced the ability to infer likely causal variants in this region using population-scale data. This novel strategy facilitated the most comprehensive evaluation to date of the role of variants in the CRP gene governing CRP levels, providing important information on the minimal subset SNPs necessary for comprehensive evaluation of the likely causal relevance of elevated CRP levels for coronary heart disease risk by Mendelian randomisation. The method is implemented in WinBUGS and could be applied to evidence synthesis of genetic association studies of other quantitative traits (or disease outcomes), whenever the typed SNPs vary among studies.

Participants

Name	Institution
Almasy, Laura	Southwest Foundation for Biomedical Research (SFBR)
Anderson, Amy	University of Washington, Seattle
Arjas, Elja	University of Helsinki
Balding, David	University College London
Baur, Max	University of Bonn
Bickeboeller, Heike	University of Goettingen
Bishop, Tim	Cancer Research UK, University of Leeds
Clayton, David	University of Cambridge
Clerget-Darpoux, Françoise	INSERM & Paris-Sud University
Cordell, Heather	Newcastle University
Didelez, Vanessa	University College London
Egeland, Thore	Ullevål University Hospital
Elston, Robert	Case Western Reserve University
Fisher, Sheila	Kings College London School of Medicine
Gilks, Wally	University of Leeds
Heath, Simon	National Genotyping Centre, France
Hepler, Amanda	University College London
Holmans, Peter	University of Wales College of Medicine, Cardiff
Lewis, Sarah	University of Bristol
McKnight, Amy Jayne	Queens' University Belfast
Nothnagel, Michael	University of Kiel
Nsengimana, Jeremie	University of Leeds
Palmgren, Juni	Karolinska Institutet, Stockholm
Plagnol, Vincent	University of Cambridge
Ripatti, Samuli	Karolinska Institute
Sheehan, Nuala	University of Leicester
Strauch, Konstantin	Philipps University Marburg
Teare, Dawn	University of Sheffield
Thomas, Alun	University of Utah
Thomas, Duncan	University of Southern California, Los Angeles
Thompson, Elizabeth A	University of Washington
Timpson, Nicholas	University of Oxford
van Duijn, Cornelia M	Erasmus University Medical Centre
Whittaker, John	London School of Hygiene and Tropical Medicine