Pairwise Relatedness Estimation: Accounting for Evolutionary Effects

1
Pairwise Relatedness Estimation Accounting for
Evolutionary Effects

• 6th Annual DNA Grantees Workshop
• Washington, DC
• June 27 - 29, 2005
• Amanda B. Hepler
• Bruce S. Weir
• NIJ Grant 2004-90398-NC-IJ

2
Outline

• What is Pairwise Relatedness?
• Why is it Important?
• What are Evolutionary Effects?
• New Maximum Likelihood Estimator
• A Simulation Study
• FBI Data Analysis

3
What is Pairwise Relatedness (PR)?

• When considering two individuals, there exists an
  amount of relatedness between them.
• For example, if they are siblings they will have
  a higher amount of relatedness than if they were
  not siblings.
• Inference about relatedness between two
  individuals can be achieved by observing
  similarities and differences between genotypes at
  various loci.
• Similarity in genetic makeup is an indicator of
  relatedness.

4
Why is PR Important?

• Remains identification
• Cannot always rely on visual ID, dental records,
  etc.
• Courtroom
• Only relative available for DNA testing.
• Defense suggests a relative is the culprit of the
  crime
• Population genetics applications

5
Evolutionary Effects

• Evolutionary theory states that subpopulations
  exist within most human populations.
• People could tend to mate within their
  subpopulation, resulting in inbreeding.
• Inbreeding adds an additional amount of
  relatedness between random individuals within a
  subpopulation.
• This extra relatedness should be taken into
  account when estimating pairwise relatedness.

6
Measuring Pairwise Relatedness

• Coancestry coefficient (?XY) probability a
  random allele from X is identical by descent to a
  random allele from Y.
• Common ?XY values
• Unrelated 0 (Min)
• Cousins 1/16
• Full sibs 1/4
• Identical Twins 1/2 (Max)
• Jacquards coefficients ? (?1, ?2, ?3, ?4, ?5,
  ?6, ?7, ?8, ?9) See handout for more details.

7
Maximum Likelihood

• Use maximum likelihood estimation.
• First we estimate Jacquards coefficients.
• Then we use the following relationshipto
  obtain an estimate for ?XY.
• If evolutionary effects are ignored ?1 ?6
  0.
• This estimator was first proposed by Thompson,
  and is commonly used today.
• E.A. Thompson. The Estimation of Pairwise
  Relationships (1975). Annals of Human Genetics.
  39 173-188.

8
Simulation Study

• Compare two estimators of pairwise relatedness
• MLE using only 3 parameters (?1 ?6 0),
• MLE using all 9 parameters (all ?i 0).
• Investigate the effects of varying the
• amount of inbreeding,
• number of loci,
• number of alleles per locus,
• and true relationship.

9
3 Parameter Model Means
Mean Relatedness Estimates vs. Number of Loci
Inbreeding Coefficient
True Relationship
10
9 Parameter Model Means
Mean Relatedness Estimates vs. Number of Loci
Inbreeding Coefficient
True Relationship
11
FBI Data

• Data obtained from http//www.fbi.gov/hq/lab/fsc/b
  ackissu/july1999/budowle.htm
• Individuals were typed at 20 markers,
• 3 28 alleles per locus.

B.S. Weir, W.G. Hill. Estimating F-Statistics
(2002). Annual Review of Genetics. 36 721-750.
12
FBI Data Results

• Randomly selected 500 unrelated pairs from each
  group.
• Mated those pairs to produce 500 full sibling
  pairs.

13
Conclusions

• Need several loci for both estimators to perform
  well.
• More alleles is better, but not as important.
• Recommend using 9 parameter estimator for inbred
  populations.
• For non-inbred, existing methods are less biased.
• THANK YOU!!