Molecular Genetic Epidemiology A Laboratory Perspective

1st Edition

Visit the link below to download the full version of this book:

https://medipdf.com/product/molecular-genetic-epidemiology-a-laboratory-perspect
ive-1st-edition/

Click Download Now

Preface

This book covers selectively, a representative range of approaches in use in lab-

oratories engaged in molecular genetic epidemiology. Following close on the
tail of the Human Genome Sequencing Program is the study of the molecular
genetic factors which make people and their diseases different. Intense
interest is resulting in development of a bewildering array of theoretical and
laboratory approaches. This volume presents overviews of core topics and
techniques which are widely available and accessible to almost any researcher.
It focuses on the robust, tried and tested methods, ranging through statistical
planning, clinical sampling, microsatellite SNP and sequence calling and appli-
cations suitable for population studies in fields including immunogenetics,
cardiovascular disease and subject identification.
Contributors are drawn from both an MRC Co-operative Group in Genetic
Epidemiology (School of Medicine, University of Southampton, UK) and inter-
nationally in related areas of expertise and they are thanked both for respond-
ing to deadlines and patience in waiting for others to meet theirs.
Over the next decade, micro array chip and nanotechnology develop-
ments will flourish for ultra-high throughput applications, but will drive an
increasing demand for established "medium" throughput and configurable
approaches. We hope that this Volume will empower individual researchers in
molecular genetic epidemiology.

June 2001 IAN DAY

Contents

1 Mapping Genes for Common Diseases:

Statistical Planning, Power, Efficiency and Informatics
ANDY COLLINS
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Power and Sampling Considerations . . . . . . .. . .............. 2
1.3 Models of Locus Action .................................. 6
1.4 Maximum Likelihood Estimation .......................... 7
1.5 Allelic Association ................ ...................... 12
1.6 The Malecot Model for Association . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Candidate Genes for Allelic Association. . . . . . . . . . . . . . . . . . . . . 17
1.8 Significance Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.9 Meta-Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.10 Informatics ............................................ 20
1.11 Summary.............................................. 24
References ............................................. 24

2 Human DNA Sampling and Banking

EMMANUEL SPANAKIS
2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 A Brief History of DNA Research .......................... 27
2.3 Goals and Needs of Human Population Genetics ............. 29
2.4 Selection of the Source Tissue ............................ 33
2.5 A Fundamental Change in Sampling Methodology ........... 36
2.6 DNA Isolation and Purification ............................ 38
2.7 Sample Storing ......................................... 42
2.8 Sample Banking ........................................ 44
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3 Microsatellite Genotyping
CHENI KWOK and KARIN SCHMITT
3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Microsatellite Markers ................................... 57
3.3 The Genotyping Process - Experimental Considerations ....... 60
3.4 Data Analysis ........................................... 62
VIII Contents

3.5 Data Management ...................................... 68

3.6 Hardware. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.7 Future Developments .................................... 76
3.8 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
References ............................................. 80

4 Minisatellite and Microsatellite DNA Fingerprinting

PAUL G. DEBENHAM
4.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Blood Grouping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3 DNA Fingerprinting - The Discovery ...................... 88
4.4 DNA Fingerprinting - The Applications .................... 90
4.5 An Explosion of DNA Fingerprinting Sequences . . . . . . . . . . . . . . 92
4.6 Shortcomings of Multilocus Probes ........................ 93
4.7 Single Locus Probes - DNA Profiling ....................... 94
4.8 Europe and the USA Disagree ............................. 96
4.9 Lessons from the Castro Case ............................. 96
4.10 Impact of PCR ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.11 Digital DNA Typing ..................................... 101
4.12 Short Tandem Repeat Profiling ............................ 101
4.l3 The UK National DNA Database ........................... 103
4.14 STR Profiling Applications ................................ 103
4.15 The Future ............................................. 104
References ............................................. 105

5 Multiplex Polymerase Chain Reaction and Immobilized Probes:

Application to Cardiovascular Disease
SUZANNE CHENG
5.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 107
5.2 Multiplex Polymerase Chain Reaction (PCR) Methodology . . . . . 108
5.3 Sequence-Specific Oligonucleotide Probe (SSOP)
Methodology ........................................... 116
5.4 Example: Candidate Markers of CVD Risk .................. 120
5.5 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 127
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 127

6 The Special Case of HLA Genes: Detection and Resolution

of Multiple Polymorphic Sites in a Single Gene
W. MARTIN HOWELL and KATHERINE L. POOLE
6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.2 The HLA System ........................................ l32
6.3 Requirement for HLA Genotyping: Historical Perspectives ..... l34
6.4 PCR-based Approaches to HLA Genotyping ................. l36
 Contents IX

6.5 A PCR-SSOP-Based Strategy for HLA Class II Typing Using

DNA Derived from Archival Tissue Banks. . . . . . . . ............ 147
6.6 Concluding Remarks .................................... 152
References ............................................. 152

7 Microplate Array Diagonal Gel Electrophoresis (MADGE)

Methodologies: The First Five Years
IAN N.M. DAY, EMMANUEL SPANAKIS, LESLEY J. HINKS,
ANCA VOROPANOV, XAIO-HE CHEN, and SANDRA D. O'DELL
7.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 157
7.2 Microplate Array Diagonal Gel Electrophoresis (MADGE) . . . . .. 161
7.3 CpG-PCR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 163
7.4 Temporal Thermal Ramp MADGE Electrophoresis
(Melt-MADGE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 165
7.5 Software Applications (Applies to All MADGE Not Just
Melt-MADGE) .......................................... 170
7.6 Future Developments .................................... 172
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 179

8 The Use of Sequence Analysis for Homozygote

and Heterozygote Base Variation Discovery
HANS-ULRICH THOMANN, MICHAEL FITZGERALD,
HEIDI GIESE, and KRISTEN WALL
8.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 183
8.2 Sequence Analysis: a Polymorphism Discovery Method ....... 183
8.3 Key Advancements in Sequencing Technologies:
Chemistries, Hardware, and Software. . . . . . . . . . . . . . . . . . . . . . .. 184
8.4 Heterozygote and Homozygote Polymorphism Base Calling .... 191
8.5 Strategies for Sequencing-Based Polymorphism Detection ..... 194
8.6 Summary and Outlook: To Sequence or Not To Sequence -
This Is Not in Question .................................. 203
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 207

Subject Index 211

List of Contributors

XAIO-HE CHEN
Human Genetics Research Division, Duthie Building (MP 808), Southampton
General Hospital, Tremona Road, Southampton S016 6YD, UK
e-mail: [email protected]
SUZANNE CHENG
Department of Human Genetics, Roche Molecular Systems, Inc., 1145
Atlantic Avenue, Alameda, California 94501-1145, USA
e-mail: [email protected]
ANDY COLLINS
Human Genetics Research Division, Duthie Building University of
Southampton, Southampton General Hospital, Tremona Road, Southampton
S016 6YD, UK
e-mail: [email protected]
IAN N.M. DAY
Human Genetics Research Division, Duthie Building (MP 808), Southampton
General Hospital, Tremona Road, Southampton S016 6YD, UK
e-mail: [email protected]
PAUL G. DEBENHAM
Laboratory of the Government Chemist, LGC Building, Queens Road,
Teddington, Middlesex TW11 OLY, UK
e-mail: [email protected]
MICHAEL FITZGERALD
Genome Therapeutics Corporation, 100 Beaver Street, Waltham,
Massachusetts 02454, USA
HEIDI GIESE
Genome Therapeutics Corporation, 100 Beaver Street, Waltham,
Massachusetts 02454, USA
LESLEY J. HINKS
Human Genetics Research Division, Duthie Building (MP 808), Southampton
General Hospital, Tremona Road, Southampton S016 6YD, UK
e-mail: [email protected]
XII List of Contributors

W. MARTIN HOWELL
Wessex Immunology Service, Southampton General Hospital, Tenovus
Laboratory, Tremona Road, Southampton University Hospitals NHS Trust,
Southampton S016 6YD, UK
e-mail: [email protected]. uk

CHENI KWOK
Exelixis Pharmaceuticals, Inc., 170 Harbor Way, P.O. Box 511,
South San Francisco, California 94083-0511, USA
e-mail: [email protected]

SANDRA D. O'DELL
Human Genetics Research Division, Duthie Building (MP 808), Southampton
General Hospital, Tremona Road, Southampton S016 6YD, UK
e-mail: [email protected]

KATHERINE L. POOLE
Combined Laboratory, Derriford Hospital, Plymouth PL6 8DH, UK
e-mail: [email protected]

KARIN SCHMITT
Exelixis Pharmaceuticals, Inc., 260 Littlefield Avenue, South San Francisco,
California 94080, USA
e-mail: [email protected]

EMMANUEL SPANAKIS
Human Genetics Research Division, Duthie Building (MP 808), Southampton
General Hospital, Tremona Road, Southampton S016 6YD, UK
e-mail: [email protected]

Current address: Aventis Pharma, Evry Genetics Center, 2, rue Gaston

Cremienx, CP 5705,91057 Evry Cedex, France
e-mail: [email protected]

HANS-ULRICH THOMANN
Genome Therapeutics Corporation, 100 Beaver Street, Waltham,
Massachusetts 02454, USA

ANCA V OROPANOV
Human Genetics Research Division, Duthie Building (MP 808), Southampton
General Hospital, Tremona Road, Southampton S016 6YD, UK
e-mail: [email protected]

KRISTEN WALL
Genome Therapeutics Corporation, 100 Beaver Street, Waltham,
Massachusetts 02454, USA
1 Mapping Genes for Common Diseases:
Statistical Planning, Power, Efficiency and Informatics

ANDY COLLINS

1.1 Introduction

After a decade or so of progress with the identification of major genes, the

emphasis has shifted towards genes for common diseases (complex traits). Real
successes have been few so far. There has been a gradual appreciation of the
difficulties presented by genes that have a relatively small individual pheno-
typic effect and which may show complex interactions. Furthermore, there
may be different genetic determinants of the disease in different populations
together with environmental effects and practical difficulties in obtaining
sufficiently large samples of suitable material for analysis. At the same time,
analytical methods have been in flux with many new tests and variations on
existing methods appearing in the literature. The choice of analytical strategy
is determined to a large extent by the nature of the disease and the DNA
resources that can be obtained. Some assessment of power to detect a genetic
locus can be useful; however, this may be of limited value given the ignorance
of the nature of the genetic basis of the disease. Consideration of the efficiency
of different approaches is also important and careful statistical planning
should be undertaken.
Figure 1.1 gives an overview of the relationship between effect on the phe-
notype (measured by a single parameter,~) and disease allele frequency. There
have been many successes with major genes (~ of 1.5 or greater), rare alleles
which have a large phenotypic effect. Examples of such single gene disorders
are the cystic fibrosis gene (CFTR) (Kerem et al.1989) and Huntington's disease
(HD) (Gusella et al. 1983). Genes of smaller individual effect (oligogenes) have
higher frequency and show complex patterns of inheritance. For polygenes the
individual effects may be so small as to make such genes undetectable with
current technology. Progress in identifying polygenes involved in complex
inheritance may come through the development of arrays of single nucleotide
polymorphisms (SNPs) in which there is a representation of the polymorphism
in every human gene (see, for example, Chakravarti 1999).

Principles and Practice

Molecular Genetic Epidemiology - A Laboratory Perspective
Ian N.M. Day (Ed.)
© Springer-Verlag Berlin Heidelberg 2002
2 Andy Collins

polygenes

oligo genes

/
major genes
/

0.1 1.5
effect J3
Fig.l.l. Relationship between disease allele frequency and phenotypic effect (p) for three gene
categories

1.2 Power and Sampling Considerations

For diseases with a Mendelian pattern of inheritance, power calculations giving

information about error rates and sample size requirements can be relatively
straightforward. For common diseases with a complex mode of inheritance
this is no longer true. In this case examination of power usually involves
estimating the sample sizes required to detect a gene that requires simplifying
assumptions about the (unknown) underlying mode of inheritance.
There are two basic approaches to identification of the oligogenes involved
in complex traits. The first is genetic linkage, which refers to the tendency of
alleles at a disease and a marker locus to segregate together within a family if
they are located physically close to each other on a chromosome. The second
approach is allelic association (also called linkage disequilibrium), which is the
tendency of a specific marker allele or alleles to be associated with a disease
allele, which may be due to the close proximity of a disease and marker locus.

1.2.1 Linkage

For major genes, complex segregation analysis, which involves fitting a para-
metric model, has been used successfully to estimate genetic parameters such
 Chap. 1 Mapping Genes for Common Diseases 3

as gene frequency, dominance and penetrances (see, for example, Morton et al.
1991). Estimates from parametric models may be used in programs such as
LINKAGE (Lathrop et al. 1984) to obtain a logarithm of odds (lod) score for
linkage of markers to a disease gene. For complex traits, segregation analysis
is faced with insurmountable difficulties. Alternatives that have been proposed
include carrying out linkage analysis under several simple models in the hope
that a single gene might have an effect large enough to be detected even under
the incorrect model. However, "nonparametric" or, more correctly, weakly
parametric approaches are usually employed to avoid fitting inappropriate
multi-parameter models.
Lio and Morton (1997) evaluated both weakly parametric models and a
parametric model, using the data presented by Davies et al. (1994), for insulin
dependent diabetes (IDDM). In this study a sample of 95 sib pairs affected
with IDDM were typed for 28 markers on chromosome 6. A number of non-
parametric methods were used including the BETA (~) model (Morton 1996),
Genehunter (Kruglyak et al. 1996) and MAPMAKERISIBS (MIS) (Kryglyak et
al.1995). The COMDS program (Morton et al.1991), which performs combined
segregation and (single locus) linkage analysis, was used as a parametric model
for comparison. The authors recognised that the data are very unsuitable for
segregation analysis (with, for example, all normal sibs omitted). However, the
results, in terms of the linkage evidence, are of some interest. Unlike the weakly
parametric approaches the COMDS program is implemented only for single
point analysis (single marker locus) and so is not able to obtain a multipoint
lod score. Multipoint mapping is more powerful as it uses genotypic infor-
mation from multiple markers in the map. Amongst the single point tests,
however, COMDS gave the highest totallod and these lods were identical or
higher to those obtained under the BETA model for markers closest to the three
putative IDDM loci (IDDM 1,5 and 8; Table 1.1).
The equivalence of combined segregation and linkage analysis to the
weakly parametric approaches seems remarkable given the inadequacy of
the ascertainment correction and the incomplete data consisting only of
affected sib pairs with parents. However, it is unlikely that the equiv-
alence would hold in all cases, motivating the use of the weakly parametric
alternatives.
Risch (1990) developed the use of the ratio A of the risk to relatives of
a proband versus the population prevalence as an indication of mode of

Table 1.1. Single marker analysis of IDDM (lads)

Marker Gene COMDS BETA MIS GENEHUNTER

D6S258 IDDMI 8.41 8.41 7.27 5.14

ESR IDDMS 1.82 1.82 1.77 1.48
D6S264 IDDM8 1.02 0.59 0.50 0.35
Total (28 markers) 52.24 49.14 44.54 32.13
4 Andy Collins

Table 1.2. Approximate minimum As for 90% power to detect

linkage

Number of affected sib pairs

60 7
70 5
80 4
200 2

inheritance. The risk ratio AR for a type R relative of an affected individual is

AR = KR/K, where K is the population prevalence and KR is the recurrence risk
for a relative of an affected individual. For example, As = 3.0 for a disease indi-
cates that siblings of a proband have a three-fold increased risk of the same
disease compared to the risk to individuals in the general population. As an
example, a disease in which 10% of siblings of affected individuals are them-
selves affected with a disease for which the prevalence in the general popula-
tion is 1/500, the value of As is 0.1/0.002 = 50.0. The risk ratio decreases with
the degree of relationship between the proband and relatives and the rate of
decrease depends on the mode of inheritance.
For linkage analysis pairs of affected relatives with different degrees of
relationship can be included using the identity by descent (IBD) between them.
Identity by descent refers to the sharing of identical ancestral alleles between
a pair of relatives (not just the sharing of an identical allele). The power to
detect linkage depends on the risk ratio Acharacterising the trait and also the
recombination fraction (8). When the risk to first-degree relatives is high (for
example AR > 5), the recurrence risk pattern may be consistent with single locus
inheritance. However, lower risk ratios suggest the presence of multiple loci
influencing the phenotype. For sib pairs, the relationship between power and
As for a range of samples can be demonstrated. For a As of 4, a sample size of
approximately 80 affected sib pairs or greater is required to detect linkage with
90% power. Table 1.2 gives the approximate number of sib pairs required for a
range of risk ratios.
This, however, is the minimum requirement as it assumes a fully informa-
tive marker and a recombination fraction between marker and disease gene of
zero. Risch also examined the influence of increasing recombination and
demonstrated that reduction in power was proportionally greater in small
samples.
The question of nonindependence of sib pairs formed in larger sibships has
received some attention. For families with multiple (s) affected or phenotyped
siblings the number of possible sib pair comparisons is s(s - 1)12. However, a
sibship of this size contributes only s - 1 independent sib pairs, but all pairs
are required for an efficient test. To compensate for this, Suarez and Hodge
(1979) proposed to count all pairs but then divide each count by sl2 so that
each pair carries weight as if there were only s - 1 pairs present. Wilson and
Fig. 1.2. The equivalent number of sib pairs compared to the actual number weighting pairs by
2/S where S is the number of siblings in the sibship and weighting all pairs equally (W = 1), for
a range of samples with varying sibship sizes

Elston (1993), considering a quantitative trait, suggested that each sib pair be
given equal weight to increase the power of the test, but the value of s - 1 be
used to determine the number of degrees of freedom contributed by a sibship
of size s. They did not show, however, that equal weights would increase power
or give the predicted distribution. Collins and Morton (1995) examined the
correlation (r) between identity by descent and a quantitative trait using data
simulated under the null hypothesis of no linkage. They considered both equal
weights (w = 1) and unequal weights for each sib pair (w = 2/S). The Fisher
Z(r) transform was used allowing evaluation of the equivalent number of sib
pairs (giving an indication of power) under both weighting schemes (Fig. l.2).
The figure shows that unequal weights per pair lose power as measured by
the equivalent number of pairs. Therefore, for complex traits, the assumption
of independence for all sibling pairs seems reasonable on the basis of available
evidence.
The majority of complex phenotypic information is on a continuous or
quantitative scale. Loci that influence a quantitative trait are termed QTLs.
Examples of such quantitative phenotypes include various asthma measures,
body mass index, and blood pressure. The power to detect QTLs depends on
6 Andy Collins

Table 1.3. Number of sib pairs to detect a QTL under an additive

model and a heritability of lO% assuming 80% power and using
the top and bottom lO% of a quantitative phenotype

Allele frequencies N pairs Number of siblings to be

screened to obtain the sample

O.lO 1,647 19,120

0.30 1,482 17,357

the heritability (h2), or the degree to which the trait is genetically determined,
and the proportion of this heritability that can be attributed to an individual
QTL. Approaches to quantitative traits are particularly sensitive to the sam-
pling scheme and the utility of extreme discordant (very different) and
extreme concordant (very similar) sib pairs has been emphasised. In particu-
lar, a sample comprising extreme discordant sib pairs from the tails of the
quantitative distribution is known to be the most powerful strategy (Risch and
Zhang 1995). Such pairs should share very few alleles identical by descent at
the trait locus. With this sampling strategy a locus accounting for only 10% of
the variance with an allele frequency of 0.3 can be detected with 1482 sib pairs
(Table 1.3). Unlike concordant pairs the discordant sib pair approach is less
susceptible to confounding due to other sources of sibling phenotype correla-
tion (including shared environment). The practical difficulties of collecting
such "ideal" samples are more of an issue, however. From Table 1.3 upwards of
17,000 siblings would have to be screened to obtain the desired number of
discordant pairs fitting the ascertainment criteria. Sampling concordant pairs
is less effective but such pairs are generally easier to obtain.

1.3 Models of Locus Action

Risch (1990) described both additive and multiplicative models of gene action.
In the case of two loci if the genotypes at the first locus are Gi, i = 1, n, and at
the second locus Hj, j = 1, m, then Wij defines the penetrance of the two locus
genotype Gi Hj. In a multiplicative or epistatic model the penetrance Wij can be
regarded as the product of two "penetrance factors" Wij = Xi Yj, implying inter-
action between genotypes at the two loci. The relationship to familial risk is
AR = ARI AR2 where 1 and 2 denote the two disease loci. Therefore, for a multi-
plicative model, the risk ratio AR is the product of "risk ratio factors" defined
through penetrance factors for the two contributing loci.
In an additive model the penetrance Wij can be regarded as a sum of
two "penetrance summands" Wij = Xi + yj. This model may appear unrealistic
implying for certain combinations of genotypes penetrances greater than
 Chap. 1 Mapping Genes for Common Diseases 7

1. This can, however, be a good approximation to genetic heterogeneity where

multiple loci contribute to a disease state but each locus is sufficient to cause
disease on its own. In terms of familial risk, the total AR - 1 is a weighted sum
of similar terms for each contributing locus. AR - 1 = ARI - 1 + AR2 - 1.
If recurrence ratios of AR for a number of types of relationship are available
the data can be used to estimate the number of loci involved in a complex
disease and suggest whether the mode of action is additive or multiplicative.
Risch (1990) performed an analysis of schizophrenia and concluded that the
observed familial risks are consistent with a model of several loci acting epista-
tically. The best fit was a single major locus with Al = 3 with a polygenic back-
ground of a large number of loci with small effects or two major loci each with
Al = 2 and a polygenic background. This finding gives some hope that one or
two determinants of larger effect (contributing most to the overall familial
association) might be detectable with current methods.

1.4 Maximum Likelihood Estimation

Typically, marker genotype data is unavailable on one or more sibs and IBD
status cannot be fully determined. For this reason maximum likelihood
methods are employed. However, where the IBD status is fully determined the
method for obtaining the lod can be demonstrated using the data of Cox and
Spielman (1989), which gives sharing proportions at HLA for sib pairs affected
with type 1 diabetes (Table 1.4).
For Table 1.4 the allele sharing proportions for (0, 1 and 2 alleles IBD) are
Zo = 10/137 = 0.073
ZI = 46/137 = 0.336
Z2 = 811137 = 0.591
and the corresponding lod is therefore
0.073 10 0.336 46 0.591 81
lod = 10gIO 10 46 8 = 16.978
0.25 0.5 0.25 I

Table 1.4. HLA sharing in sib pairs with type 1 diabetes (IDDM)

Alleles IBD

o 2 N pairs

Observed 10 46 81 137
Expected 34 69 34