2011 American Fisheries Society National Meeting presentation. Comparison of different techniques for identifying informative SNPs. The published manuscript "Rank and Order: Evaluating the Performance of SNPs for individual assignment in a non-model organism" is available open access through PLoS One.

1. RANK AND ORDER:
EVALUATING THE PERFORMANCE
OF SNP ASSAYS FOR SOCKEYE
SALMON
American Fisheries Society Annual Meeting 2011
Caroline G. Storer1, Carita E. Pascal1, Steven B. Roberts1, William D. Templin2, Lisa W.
Seeb1, and James E. Seeb1
1School of Aquatic and Fishery Sciences, University of Washington, Seattle, Washington 98195, USA
2Gene Conservation Laboratory, Division of Commercial Fisheries, Alaska Department of Fish and Game, Anchorage, Alaska
99518, USA

3. http://grizzlybearblog.wordpress.com/category/grizzly-bears-fishing/

5. MICHAEL MELFORD/National Geographic Stock

6. MICHAEL MELFORD/National Geographic Stock

7. Applying genetics to fisheries
management

8. Applying genetics to fisheries
management
 Stock identification
 Inferring population history
 Parentage analysis
 Fisheries forensics
 Estimating mixed stock
composition

9. Applying genetics to fisheries
management
 Stock identification
 Inferring population history
 Parentage analysis
 Fisheries forensics
 Estimating mixed stock
composition
Ackerman et al. 2011
Genetically similar spawning populations

10. Molecular Markers, Today

11. Molecular Markers, Today
Single nucleotide polymorphisims (SNPs)

12. Molecular Markers, Today
Single nucleotide polymorphisims (SNPs)
Population 1
ACTC G
Population 2
ACAC G
SNP locus

13. Molecular Markers, Today
Single nucleotide polymorphisims (SNPs)
Population 1  Abundant
ACTC G
 The number of available
markers is growing
Population 2
 Not all SNPs are equal
ACAC G
SNP locus

14. Objectives
1)Rank existing SNPs in sockeye salmon for
performance
1)Evaluate ranking methods for SNP panel
selection

15. Approach
1)Measure genetic variation at 114 SNP loci
1)Rank loci according to five different measures of
performance
1)Evaluate ranking measures for developing 48-
and 96-SNP panels

16. The populations
South-central
Bristol Bay Alaska
Russia British
Alaska Peninsula
Columbia
Washington
Genotyped 12 populations, 61- 93 fish per population, representing 6 regions

17. Range wide genetic variation
Principal Coordinate 1 (44.5%) Bristol Bay
Russia
Alaska Peninsula
South-central
Alaska
British
Columbia
Washington
Principal Coordinate 2 (15.5%)

18. Bristol Bay
Alaska Peninsula
South-central Alaska
Russia
Washington
British Columbia

19. Ranking approaches
 FST (Weir & Cockerham 1984)
 Average contribution of each locus to principal
components (LC)
 Informativeness of assignment (In; Rosenberg et al.
2003)
 BELS (Bromaghin 2008)
 WHICHLOCI (Banks et al. 2003)

20. Locus ranks
Locus rank
Loci ordered by average rank

21. Locus ranks
Locus rank
top ranked loci
Loci ordered by average rank

22. Testing panel performance
 Created 48- and 96-SNP panels containing top ranked loci
for each approach

23. 96-SNP panel
Locus rank
Loci ordered by average rank

24. 48-SNP panel
Locus rank
Loci ordered by average rank

25. Testing panel performance
 Created 48- and 96-SNP panels containing top ranked loci
for each approach
 fORCA simulation of individual assignment (Rosenberg 2005)
- Uses the allele frequencies for user described populations to assign a
simulated individual back to the correct population
- Reports the probability that this assignment is correct

26. 96-SNP panel performance
FST In LC B S I R
W
fORCA score
E H a
L I n
C d
H o
L m
O
C

27. 48-SNP panel performance
W L
H O
FST In LC I C
B LC I
R d
E SH
a o
n m
fORCA score

28. SNP panel performance
96 SNPs 48 SNPs
fORCA score
FST In LC B LW LR d
E SH Oa o
I Cn m
C I
H

29. Results
 Most loci were ranked differently using each method
 96-SNP panels contained many of the same loci
with only 3-7 loci differing between panels
 96-SNP panels performed similarly better then 48-
SNP panels
 FST, In, and LC 48-SNP panels performed similarly
and significantly better than the BELLS and
WHICHLOCI panels

30. Lessons learned
 More loci is often better
 The more loci to choose from the more important
the ranking approach
 Be wary of upward bias

31. Next Steps
Determining the value of a locus for its resolving power
versus its value due to uniqueness of information
High
resolving
power
Correlation of loci with first two principal componants

32. Next Steps
Determining the value of a locus for its resolving power
versus its value due to uniqueness of information
Unique
information
High
resolving
power
Correlation of loci with first two principal components

33. Acknowledgements
 Alaska Sustainable Salmon Fund
 Bristol Bay Regional Seafood Development Group
 The Gordon and Betty Moore Foundation
 The Seeb Lab
 The School of Aquatic and Fishery Sciences
 AFS Genetics Section