1. DIVERSIFYING SELECTION AND FUNCTIONAL CONSTRAINT
ESTIMATING THE dN/dS RATIO FOR GENE SEQUENCES IN THE PRESENCE OF RECOMBINATION
Danny Wilson
12th October 2004

2. Menu Codon-based models of molecular evolution
An new method for estimating omega with recombination
Does it work? Simulation studies and example data

3. Codon-based models of molecular evolution
Part one
Codon-based models of molecular evolution

4. Sampling usually occurs at this point i.e. post-selection
Ancestral type
Neutral mutant
Inviable mutant
Sampling usually occurs at this point i.e. post-selection
Mutation
Selection
Underlying rates of non-synonymous mutation are usually confounded with selection against inviable mutants. Thus it is convenient to model functional constraint as mutational bias. (Or rather, make no attempt to disentangle the two).

6. Types of single nucleotide mutation Transitions vs. transversions
Purine
Transitions
Transversions T C
Pyramidine
Transitions
For any base there are always 2 possible transversions and 1 possible transition.

7. Types of codon mutation Synonymous vs. non-synonymousG A
Leucine T G A
Leucine
Methionine
Synonymous
Non-synonymous
Leucine
pH 5.98
6-fold degeneracy in the genetic code
Methionine
pH 5.74
Single unique codon ATG
CH3-S-(CH2)2-CH(NH2)-COOH
(CH3)2-CH-CH2-CH(NH2)-COOH

8. Example: CTT C T T T T T A T T G T T T C T T A T T G T T T C T T A T T
Phe
Non-synonymous transition
wkm
Ile
Non-synonymous transversion wm
Val
Ser
Tyr
Cys
Leu
Synonymous transversion m km A T T
Leucine G T T T C T T A T T G T T T C T T A T T G

9. Nielsen and Yang (1998) codon-based model of molecular evolution
Mutation rate
Synonymous transversion m
Synonymous transition km
Non-synonymous transversion wm
Non-synonymous transition
wkm
Other
Interpretation k
Transition-transversion ratio
w = dN/dS
Relative viability of non-synonymous mutations

10. codeML
Pros
Viable method for detecting mode of selection on a codon sequence
Cons
Categorizes possible values for omega into a small number of discrete intervals
Results can be misleading in the presence of recombination

11. An new method for estimating omega with recombination
Part two
An new method for estimating omega with recombination

12. Inference with recombination

13. Li and Stephens (2003) Approximation to the likelihood

14. Li and Stephens (2003) Approximation to the likelihood
TTTGATACTGTTGCCGAAGGTTTGGGCGAAATTCGCGATTTATTGCGCCGTTATCATCAT
TTTGATACCGTTGCCGAAGGTTTGGGTGAAATTCGCGATTTATTGCGCCGTTACCACCGC
TTTGATACCGTTGCCGAAGGTTTGGGTAAAATTCGCGATTTATTGCGCCGTTACCACCGC
TTTGATACCGTTGCCGAAGGTTTGGGCGAAATTCGTGATTTATTGCGCCGTTATCATCAT

15. Li and Stephens (2003) Approximation to the likelihood
TTTGATACTGTTGCCGAAGGTTTGGGCGAAATTCGCGATTTATTGCGCCGTTATCATCAT
TTTGATACCGTTGCCGAAGGTTTGGGTGAAATTCGCGATTTATTGCGCCGTTACCACCGC
TTTGATACCGTTGCCGAAGGTTTGGGTAAAATTCGCGATTTATTGCGCCGTTACCACCGC

16. My modification to Li and Stephens(2003)

17. Estimating variable omega
The problem
A constant omega model is prone to averaging positive and negative omegas in a gene
Allowing every site its own omega leaves little information for inference
The solution
A change-point model where windows of adjacent sites share the same omega

18. Estimating variable omega
MCMC moves:
Change omega for a single block
Extend a block 5’ or 3’
Split an existing block
Merge adjacent blocks w1 w2 w3 w4 w5

19. Does it work? Simulation studies and example data
Part three
Does it work? Simulation studies and example data

20. Posterior distribution for known and unknown genealogy

21. Posterior distribution for known and unknown genealogy

22. Neutral dataset
True omega
Posterior mean
Posterior HPD interval

23. Non-neutral dataset
True omega
Posterior mean
Posterior HPD interval

24. HIV envelope gene Slow Non-Progressors vs Rapid Progressors

25. HIV envelope gene Slow Non-Progressors vs Rapid Progressors

26. Neisseria meningitidis PorB3

27. Neisseria meningitidis PorB3
95% HPD Upper
95% HPD Lower

28. Work in progress… Variable recombination rate Model indels
Falsifiability test
Test for sensitivity to rate heterogeneity

29. Acknowledgements Gil McVean (Supervisor) Martin Maiden (Supervisor)
Ziheng Yang
Rachel Urwin (meninge data)
Charlie Edwards (HIV data)