Dissecting the Genetic Complexity of the HLA Association With

							 Modeling Complexity, the Elston
Pedigree and Genetic Epidemiology


  Chris Amos, U.T. M.D. Anderson
          Cancer Center
           Thanks to Sponsors!
•   Triaj, Inc.
•   Glaxo SmithKline
•   John Wiley and Sons, Ltd.
•   Statistical Solutions, Ltd.
•   Institutional Sponsors
    – Case-Western Reserve University
    – Washington University, St. Louis
• Louisiana State University Medical Center
    Thanks to Local Organizers
•   Bronya Keats, Ph.D. - Chair
•   Stephanie Laurent
•   Diptasri Mandal, Ph.D.
•   Judy Laborde
•   Jerlyn Rose

• Jean MacCluer, Ph.D.
• Vanessa Olmo
 Thanks to Scientific Organizing
           Committee
• John Witte (Chair)
• Jenny Chang-Claude
• Bronya Keats
• Kim Siegmund
• Cornelia Van Duijn
• Bruce Weir
        Genetic Epidemiology
• "the study of genetic components in
  complex biological phenomena" JVNeel
• The successful mapping of the human
  genome will greatly facilitate the study of
  health and disease in a manner that
  integrates both host and environmental
  factors.
  Why Identify Genetic Factors for
       Complex Diseases?
– Some genetic factors may permit modulation

– Identifying and conditioning on genetic factors
  may assist in identifying environmental triggers
Alice laughed: "There's no use trying," she said; "one
can't believe impossible things."
"I daresay you haven't had much practice," said the
Queen. "When I was younger, I always did it for half
an hour a day. Why, sometimes I've believed as many
as six impossible things before breakfast."
Robert Elston, Vanquisher of Chaos
 Algorithmic and computational
    approaches to modeling
          complexity
• Elston RC. Simulation d‟un processus
  stochastique implique dans la gestion d‟une
  banque de sang. Biometrie-prazimetrie
  3:129-140, 1962.
• Elston RC and Pickrel JC. A statistical
  approach to ordering and usage policies for
  a hospital blood bank, Transfusion 3:41-47
  Summary (Elston and Pickrel,
            1963)
• An electronic computer (Univac 1105) is used
  in a simulation study to determine ordering and
  usage policies for a hospital blood bank. It is
  assumed that only type specific transfusions
  are given, and that all use of fresh blood ob-
  tained for specific cases is excluded. A ratio
  of 25:4 is taken as the ratio of the cost when-
  ever the bank is “short” of a unit for actual
  transfusion to the cost of an outdated unit.
     Elston and Stewart, 1971
• Recognition that extended families
  contain (far) more information for
  evaluating segregation-related
  parameters than smaller families
  Elston RC, Stewart J. A General Model for
 the Genetic Analysis of Pedigree Data. Hum
                Hered 21:523-534.
• Provides algorithms for estimating single
   and multiple major gene effects (possibly
   linked) in extended pedigrees
• Algorithm for polygenic effects in pedigrees
• Discusses the mixed model
• Provides methods for genotype
   classification (genetic counseling)
• Provides approaches for hypothesis testing
   (randomization, Wald tests, LR tests)
     Stewart‟s comments:
  Human Heredity 1992; 42:915
• „When I first met Robert Elston in 1968 and he
  introduced me to maximum likelihood techniques,
  I rapidly formed the conviction that these methods
  held the key to a canonical solution to a problem I
  had been wrestling with: the genetic analysis of
  characters which do not display an immediate one-
  to-one correspondance between discrete
  phenotypes and Mendelian genotypes.. . .I
  predicted that it would become a landmark paper.
  Robert was sceptical.‟
    Stewart‟s comments (cont)
• „In order to interpret a diversified syndrome
  of symptoms in terms of a branching tree of
  causes and effects from a single initial cause
  (the segregating allele), considerable
  biological knowledge is necessary.‟
• Stewart comments on the need for
  development and study of experimental
  organisms to understand biological effects
  of genes, holistic not reductionist approach
   Genetic Analysis of Familial
    Periodic Fever Syndromes
• Autosomal Dominant
• Variable phenotype consisting of
  episodic/periodic high fevers sometimes
  requiring long-term treatment with steroids
• Extended Pedigrees, Age-Dependent
  Penetrance, requires Elston-Stewart
  Algorithm
• Families collected from registries, can‟t
  correct for ascertainment
Genetic Linkage analysis of 3 Familial
Periodic Fever Families
              4
                                                               Legend
                                                    All Families  Family B
                                                    Family A      Family C

              2
  LOD SCORE




              0




          -2




          -4
                  -0.4   -0.2                 0                   0.2   0.4
                           D12S364




                                              D12S356

                                              D12S314
                                                        D12S100
                           D12S356

                                     D12S77



                                              D12S99
                                              D12S93
   Salient Developments in Mapping
        Familial Periodic Fevers
• Lack of linkage to TNF receptor 1 (p55)
• Increased p55 levels from patients with
  FPF compared to family controls (bad
  assay?, no replicate on fresh sample)
• Missense mutations of p55 cause
  increased binding to cellular
  membrane (decreased serum levels)
         Treatment of FPFs
• P75 (TNFR2) had previously been approved
  for use and treatment in RA patients
• P75 administration leads to a reversion of
  most clinical symptoms of FPFs
           A few comments
• Biological knowledge helped in identifying
  appropriate measures, but did not provide
  correct knowledge of the effects on p55.
  Quantitative analysis was critical.
• Need for robust approach in evaluating
  pedigree information, multiple potential
  sources for error (clinical mistyping).
• Need for international collaborative study to
  attain sufficient power for disease mapping
“Brother Mendel! We grow tired of peas!”
  Models for Quantitative Traits
• Numerous alleles may affect the trait levels,
  presenting a problem in building an
  adequate model
• Distributional assumptions may play a role
  in inferences (hard to evaluate whether
  nonnormality is due to skewing or allelic
  effects)
• Multiple loci are likely to be involved
   Development of Quantitative
      Modeling Approach
• Haseman-Elston (1972) approach provides
  a rapid, efficient, robust test for linkage
  using quantitative traits
• Numerous extensions: eg. Blackwelder and
  Elston, 1982; Amos and Elston, 1989
• Extension to covariance modeling (Elston et
  al. 2000), multivariate modeling
• Very wide application - e.g. mapping of
  DBH (Wilson et al., 1988)
         Other Contributions
• Models for multilocus segregation analysis
  in backcrosses and intercrosses
• Methods for ascertainment correction
• Cost and efficiency of genetic linkage
  procedures

• Huge number of students
         The Elston Pedigree
• We (naively) set out to collect data on the
  „offspring‟ of Robert Elston.
• Pedigree Spans 5 generations and includes
  382 (possibly redundant) individuals
• Limited inbreeding - indicating a preference
  for exogamous relationships with some
  exceptions (inbreeding at UCLA)
• All talks in the first session have a student
  of Elston‟s among the authors, great breadth
UNC Chapel Hill (60-79)
LSU New Orleans (79-95)
Elston Pedigree - CWRU
   Elston Pedigree - Comments
• Some Students had very large clutches,
  Anne Spence, Nancy Mendell, Joan Bailey-
  Wilson, Lynn Goldin. Some overlaps.
• Pedigree fails to show the web of complex
  relationships to Elston
• Trainees come from very wide range of
  backgrounds, disciplines
       Robert Elston and IGES
• Strong impact on the I and S parts of the
  society
• Strong emphasis on the International
  conduct of science
• These international ties are critical but must
  be nurtured following 9/11. Travel
  restrictions limit travel of US alien residents
  but also impede travel from other countries
  Robert Elston and the Society
• First Secretary of Society 1993-1994, spent
  hours struggling with rules for incorporation
  in Louisiana
• With DC Rao drafted the By-Laws of the
  society
• Served as President 1997
     Societal Issues in Genetic
          Epidemiology
• Increasing pressure from IRBs and other
  regulatory bodies impede collection of
  patient information, sharing of data (even
  after some level of de-identification)
• HIPAA regulations impede the abstraction
  of medical chart data, sharing of patient
  information from hospitals, affect informed
  consent process
Role of IGES in the International
      Genetics Community
• Limited membership with a focused area of
  research - focused areas of expertise
• Associate member of the International
  Federation of Human Genetics Societies
• Widely read manuscript reviewing charges
  from Tierney against conduct by James V.
  Neel (Thanks to Duncan, Max, Partha)
• Outreach through Education committee to
  Epidemiological and Genetic Societies
   Identifying genetic factors for
    complex traits and diseases
• Most effective approaches depend upon
  situation (which is probably unknown)
• Genetic Model-free methods are easily
  applied and can allow for oligogenic
  inheritance will little information loss
   – Goring H, et al. Genet Epidemiol 2001;21 1:S783
• Monte Carlo Markov Chain approaches can be
  effective for resolving multiple genetic effects, but
  operating characteristics for overdetermined
  models need further study
   Resolving Complex Genetic
             Effects
• Newer machine-learning tools hold promise
  for gene identification
  – Moore JH, Hahn LW Pac Sym Biocomput 2002;:53-64

  – Shannon WD et al. Genetic Epidemiology 2001,
    20:293-306.
• Cladograms and other approaches for
  sorting data show promise for interpreting
  complex associations
Tree Diagram 8: Chr1, 269 cM
                                        Both sibs have 2 DR4 alleles
                          =0.54

                              Non-Caucasians
            =0.47                              =0.55

                                               Concordant
   A   T
                                  =0.54       for nodules       =0.65
  G     C
  C     G
   T   A

                     =0.51                                   sib with RF>100
                                                              1
                                               =0.56
   A   T
  G     C
  C     G
                                                                          Both sibs have 0 DR4 alleles
   T   A                         =0.53                      =0.64

   A   T                                                                                  Same sex sibs
  G     C                                      =0.56                       =0.68
  C     G
   T   A

                                                              =0.61                         =0.71
                                                                              T22b
    Resolving Complex Effects
• Studies in isolated populations may provide
  novel insights
  – Greater extent of linkage disequilibrium
  – Less complex genetic architecture
  – More power for identifying recessive loci
• Animal models can be highly effective for
  evaluating potential genetic factors in
  humans (Carrasquillo MM et al., Nat Genet
  2002;32:237-44)
Figure a, Region analyzed from
HLA-DNA to Tap b, LD in this
region in north Europeans. Classic
|D'| measures27 of complete LD
(lower right), where |D'|=1 for
marker pairs showing only three
haplotypes, are shown for all pairs
of markers with minor allele
frequencies of at least 0.15,
together with the associated
likelihood ratio (LR) versus free
association (upper left), and color-
coded as indicated at top right. c,
Corresponding LD plot for markers
with minor allele frequencies less
than 0.15. d, Expansion of plot b in
the HLA-DNA region e, Expansion
of plot b in the HLA-DMB region.
Nat Genet 29:217-222, 2001.
     Genetic Epidemiology and
       Population Genetics
• Evaluation of haplotype structure in
  different populations indicates the value of
  population genetic studies
• Importance of clear knowledge of genetic
  background of populations indicates need
  for collaboration with
  fieldworkers/anthropologists
• Existence of LD over extended regions
  indicates value of association approaches
    Genetic Epidemiology and
         Epidemiology
• An ultimate goal in genetic epidemiology is
  identification of modifiable factors
• Environment is hard to measure but much
  easier to change than genetic factors
• Continued interaction with our
  epidemiological colleagues will help us to
  develop optimal designs for understanding
  genetic and environmental determinants of
  disease
      Studying Complex Traits
• Best studies are those with clear hypotheses
  - Often best science results from finding
  one‟s hypothesis is completely incorrect e.g.
  Williamson and Amos, 1995.
• Role for descriptive science - e.g.
  microarray discovery, but results are far
  more readily interpreted under hypotheses
     Complex Models, Genetic
     Epidemiology and Elston
• Some of Elston‟s great success results from
  an effective cross-fertilization of statistics,
  computational methods and genetics
• Current underrepresentation of animal
  methods/models (at this meeting, generally
  in our literature)
• Need for open-minded approach to new
  computational tools, many tools need
  further statistical development/ validation
          Acknowledgements
•   Monawar Hosain   -   Elston Pedigree
•   Sanjay Shete     -   Comments
•   Carol Etzel      -   Slides
•   Tracy Costello   -   Slides