Hywel B. Jones

Population structure, differential bias and genomic control in a large-scale, case-control association study

The main problems in drawing causal inferences from epidemiological case-control studies are confounding by unmeasured extraneous factors, selection bias and differential misclassification of exposure. In genetics the first of these, in the form of population structure, has dominated recent debate. Population structure explained part of the significant +11.2% inflation of test statistics we observed in an analysis of 6,322 nonsynonymous SNPs in 816 cases of type 1 diabetes and 877 population-based controls from Great Britain. The remainder of the inflation resulted from differential bias in genotype scoring between case and control DNA samples, which originated from two laboratories, causing false-positive associations. To avoid excluding SNPs and losing valuable information, we extended the genomic control method by applying a variable downweighting to each SNP.

Empirical genomewide significance levels established by whole genome simulations

The advent of high‐resolution genetic maps and semiautomated genotyping technology has opened the way for genome screening in genetically complex traits. Many such screens are now under way, or completed, most using multipoint nonparametric linkage analysis of affected sibling pairs. This type of linkage analysis uses all the available genotype information to calculate the maximum lod score (MLS) value at each point in the genome, and thereby generates MLS profiles along each chromosome. Any positive MLS values indicate potential linkage, but the peaks in these profiles, which may be referred to as “hits,” identify the most likely locations of disease susceptibility genes. However, such analysis presents serious problems of multiple testing, and the assessment of the statistical significance of hits has become a contentious issue [Lander and Kruglyak (1995) Nat Genet 11:241–247; Curtis (1996) Nat Genet 12:356–357; Witte et al. (1996) Nat Genet 12:355–356]. Having recently completed a genome screen in multiple sclerosis, we decided to investigate the statistical properties of our study by simulation. We report here in detail the results of this simulation study. Our main conclusion is that, for the particular set of families and markers used in our screen, an MLS of 3.2 carries a genome‐wide significance of 5% (that is, there is a 5% probability of observing at least one false hit, above this threshold in a complete genome screen). This value is closer to the familiar limit of 3.0, originally suggested by Morton [1955; Am J Hum Genet 7:277–318] than to the more stringent limit of 4.0 recently proposed by Lander and Kruglyak [1995; Nat Genet 11:241–247]. This is somewhat reassuring, in view of the very large sample sizes that would be necessary to achieve adequate power to detect linkage at the more stringent threshold. Genet. Epidemiol. 14:223–229,1997.

HLA typing in the United Kingdom multiple sclerosis genome screen.

ABSTRACT The United Kingdom multiple sclerosis genome screen demonstrated a peak maximum lod score of 2.8 in the HLA region, together with statistically significant excess transmission of the 121-base pair (bp) allele of the tumour necrosis factor-a marker. In order to determine whether this association is independent of the established HLA association, or simply a consequence of the 121-bp allele being part of the same haplotype, we HLA-DR and -DQ typed the 227 sibling-pair families used in the original screen. The expected associations of multiple sclerosis with the DR15 (p=8.7E-18), DQ6 (p=2.0E-09) and DR51 (p=2.8E-16) phenotypes were confirmed, and excess transmission of the DRB1*1501 and DQB1*0602 alleles was demonstrated. Combining HLA typing with the original microsatellite data demonstrated extensive linkage disequilibrium between the 121-bp allele and the 1501-0602 haplotype. Outside this extended haplotype (121-1501-0602), none of the alleles demonstrated significant transmission distortion. Having established the importance of this extended haplotype, we reanalysed the entire genome screen data after excluding those sibling pairs sharing the extended haplotype (n=27). Conditioning the full genome screen data on the basis of identity by state sharing showed that some potential linkage regions identified in the original screen clustered in families, in which the extended haplotype was shared (1p, 2p and 17q), whereas others grouped with those in which it was not (5cen, 7p and Xq). This suggests complexity in the genetics of multiple sclerosis.

Optimal genotype determination in highly multiplexed SNP data

High-throughput genotyping technologies that enable large association studies are already available. Tools for genotype determination starting from raw signal intensities need to be automated, robust, and flexible to provide optimal genotype determination given the specific requirements of a study. The key metrics describing the performance of a custom genotyping study are assay conversion, call rate, and genotype accuracy. These three metrics can be traded off against each other. Using the highly multiplexed Molecular Inversion Probe technology as an example, we describe a methodology for identifying the optimal trade-off. The methodology comprises: a robust clustering algorithm and assessment of a large number of data filter sets. The clustering algorithm allows for automatic genotype determination. Many different sets of filters are then applied to the clustered data, and performance metrics resulting from each filter set are calculated. These performance metrics relate to the power of a study and provide a framework to choose the most suitable filter set to the particular study.

A SCREEN OF CANDIDATES FROM PEAKS OF LINKAGE : EVIDENCE FOR THE INVOLVEMENT OF MYELOPEROXIDASE IN MULTIPLE SCLEROSIS

We tested 11 microsatellite markers for evidence of transmission distortion in 744 trio families with multiple sclerosis. Ten of the markers lie within or near to candidate genes selected on the basis that they map within the regions of potential linkage identified in our previously reported linkage genome screen, while the eleventh is an anonymous marker which had previously shown modest evidence for transmission distortion in our sibling pair families. Only the marker related to the myeloperoxidase (MPO) gene revealed tentative evidence for linkage disequilibrium and further work on this gene is clearly needed in order to resolve the status of this region in conferring susceptibility to multiple sclerosis.

Exploring the dense mapping of a region of potential linkage in complex disease: an example in multiple sclerosis.

In 1996 we reported the results of a genome screen in multiple sclerosis, in which potential linkage was identified in a total of twenty regions, including the centromeric region of chromosome 5. In order to investigate the efficiency of typing dense arrays of markers in regions of potential linkage, we have typed an additional nineteen microsatellite markers from this chromosome 5 region (D5S623 ‐ D5S428) in the same sibling pair families. The mean additional information extracted per marker typed declined with increasing map density, while inaccuracies in the mapping and the density of genotyping errors increased. Our empirical results suggest that, in linkage‐based experiments, there is a limit to the benefits that are gained from typing additional markers in the same families. Increasing map density up to the 2.5–5 cM level efficiently extracts valuable extra information; however, beyond this level efficiency declines while the confounding effects of mapping and genotyping errors accumulate. We, therefore, recommend that extra markers typed in linkage studies be limited to this level of resolution. Mapping regions beyond this density should only be initiated when searching for linkage disequilibrium. Genet. Epidemiol. 17:51–63, 1999.

A large-scale screen for coding variants in SERT/SLC6A4 in autism spectrum disorders.

In the current study we explored the hypothesis that rare variants in SLC6A4 contribute to autism susceptibility and to rigid‐compulsive behaviors in autism. We made use of a large number of unrelated cases with autism spectrum disorders (∼350) and controls (∼420) and screened for rare exonic variants in SLC6A4 by a high‐throughput method followed by sequencing. We observed no difference in the frequency of such variants in the two groups, irrespective of how we defined the rare variants. Furthermore, we did not observe an association of rare coding variants in SLC6A4 with rigid‐compulsive traits scores in the cases. These results do not support a significant role for rare coding variants in SLC6A4 in autism spectrum disorders, nor do they support a significant role for SLC6A4 in rigid‐compulsive traits in these disorders.

Evidence that allelic variants of the spinocerebellar ataxia type 2 gene influence susceptibility to multiple sclerosis

Expanded CAG trinucleotide repeats are known to be responsible for five of the autosomal dominant spinocerebellar ataxias (SCA1, SCA2, SCA3, SCA6, and SCA7). We have typed each of these repeats in 226 multiple sclerosis sibling pair families. No expanded repeats were seen, indicating an absence of SCA phenocopies in clinically defined familial multiple sclerosis. However, transmission disequilibrium testing for these repeats demonstrated significant excess transmission of the 22 repeat length allele of the SCA2 gene (P=4.4E-06) in multiple sclerosis patients. This observation is consistent with pleiotropic effects of the SCA2 gene, with a non-dynamic mutation/polymorphism contributing epistatically to susceptibility in multiple sclerosis.

Regional assignment of human ESTs by whole-genome radiation hybrid mapping.

The UK HGMP Resource Centre’s collection of human partial cDNA sequences (ESTs) have been examined for suitability for mapping by PCR on a panel of somatic cell hybrids. The chromosomal assignments of 92 ESTs were determined with a monochromosomal hybrid panel, and a subset of 45 were linked to genetic markers with a panel of whole-genome radiation hybrids (WG-RHs). These results demonstrate the potential of WG-RHs to construct a transcript map of the human genome.

NO EVIDENCE FOR ASSOCIATION OF MULTIPLE SCLEROSIS WITH THE COMPLEMENT FACTORS C6 AND C7

Four genome screens in multiple sclerosis have been completed and each has identified evidence for linkage in the pericentromeric region of chromosome 5. This region encodes a number of candidate genes including those for the complement components C6, C7 and C9. We have used a multiplexed oligoligation assay (OLA) to test single nucleotide polymorphisms (SNPs) from the C6 and C7 genes for evidence of association with multiple sclerosis in our sibling pair families. There was no statistically significant difference in the allele frequencies of these polymorphisms in the index cases from our families when compared with locally derived controls. No evidence for transmission distortion was seen with any of the polymorphisms, or with the haplotype built from the three SNPs from the C7 gene. Despite offering themselves as potential candidates these complement genes appear not to confer susceptibility to multiple sclerosis.