1. What is positive selection?
dN = rate of nonsynonymous
substitution
dS = rate of synonymous substitution
Let  ratio = ratio of dN/dS

2. Positive selection occurs when the  ratio exceeds unity
__________________________________
Type of Selection Outcome
Purifying selection dN/dS < 1
No selection dN/dS = 1
Positive selection dN/dS > 1

3. How do we test for positive selection?
1. Estimate means and variances of dN and dS for all pair-wise species comparisons.
2. Use t-test to determine if dN and dS differ significantly.

4. Some problems…
1. Averages over all amino acid positions in a protein.

5. Some problems… 1. Averages over all amino acid positions in a protein.
2. Averages over all lineages.

6. Some problems… 1. Averages over all amino acid positions in a protein.
2. Averages over all lineages.
3. Can detect positive selection only when it is very strong and consistent through evolutionary time.

7. Some more problems…
Ignores the phylogenetic framework in which adaptive molecular evolution occurs!

8. Suppose a significant  ratio is detected between cow and pig
1. Cow --- ns ns ns ns
2. Deer ns ns ns ns
3. Whale ns ns ns
4. Hippo ns ns
5. Pig ns
6. Camel
ns = non significant

9. A phylogenetic perspective
 > 1?
 > 1?  
Cow
 > 1? 
Deer
Whale
 > 1?
Hippo 
Pig
Camel
Outgroup

10. An example: lysozyme evolution in colobine monkeys
• Colobine monkeys are leaf-eaters that have evolved a complex foregut (like ruminants).
• Stomach expresses a high level of the bacteriolytic enzyme, lysozyme.

11. Phylogeny of Colobines and
Cercopithecines
Foregut
fermentation
evolved 
Hanuman langur
Purple-faced langur
Dusky Langur
Francois’ Langur
Proboscis monkey
Guereza colobus
Angolan colobus
Patas monkey
Vervet
Talapoin
Rhesus macaque
Allen’s monkey
Olive baboon
Sooty mangabey
Chimpanzee
Colobines
Cercopithecines
from Messier & Stewart (1997)

12. Phylogeny of Colobines and
Cercopithecines
Hanuman langur
Purple-faced langur
Dusky Langur
Francois’ Langur
Proboscis monkey
Guereza colobus
Angolan colobus
Patas monkey
Vervet
Talapoin
Rhesus macaque
Allen’s monkey
Olive baboon
Sooty mangabey
Chimpanzee
 = 4.7 
Colobines
Cercopithecines
from Messier & Stewart (1997)

13. A Maximum-Likelihood (ML) approach to the detection of positive selection

14. A Maximum-Likelihood (ML) approach to the detection of positive selection
ML methods evaluate the probability (i.e., likelihood) of obtaining a set of DNA sequences given:
a specific phylogenetic tree
an explicit model of nucleotide substitution.

15. Some details of the model…
 Implemented in the PAML package of Yang (1997)
 Uses a Markov process to describe substitutions between sense codons
 Parameters include: transition/transversion ratio ()
codon frequencies ()
branch lengths scaled for time (t)

16. Testing for positive selection involves comparing two models:
Model M7: Assumes  ratios follow a beta distribution (i.e., constrained in the interval 0-1).

17. Testing for positive selection involves comparing two models:
Model M7: Assumes  ratios follow a beta distribution (i.e., constrained in the interval 0-1).
Model M8: Adds a second class of sites to M7 at which  ratios can exceed unity (i.e., positive selection).

18. Statistical testing can be done by likelihood ratio tests (LRTs)
1. Obtain log likelihood score from model M7, ℓM7 (null model).

19. Statistical testing can be done by likelihood ratio tests (LRTs)
1. Obtain log likelihood score from model M7, ℓM7 (null model).
2. Obtain log likelihood score from model M8, ℓM8 (positive selection).

20. Statistical testing can be done by likelihood ratio tests (LRTs)
1. Obtain log likelihood score from model M7, ℓM7 (null model).
2. Obtain log likelihood score from model M8, ℓM8 (positive selection).
3. Test for significance:
X 2 = 2 (ℓM8 – ℓM7 ) with 1 d.f.

21. Advantages of ML approach
1. Allows for formal statistical testing by likelihood ratio tests.

22. Advantages of ML approach
1. Allows for formal statistical testing by likelihood ratio tests.
2. Allows for individual codons subject to positive selection to be identified.

23. Advantages of ML approach
1. Allows for formal statistical testing by likelihood ratio tests.
2. Allows for individual codons subject to positive selection to be identified.
3. Allows for positive selection to be inferred along individual branches of a phylogeny.

24. Application to the pantophysin gene in marine gadid fishes
 Pantophysin is an integral membrane protein localized to small (<100 nm) cytoplasmic microvesicles
 Believed to function in a variety of intracellular shuttling pathways
 Exact function remains unknown

25. Transmembrane structure of pantophysin
V N E E I F A S F N Y P F R L M
T S I V A L
S Q
P S
P P
S D V C
Lumen of
microvesicle
G T P V Q Y
T R
K C
T D S
Q K N G
A G
V T T E S
W N
N F
K C T
I L
V A G
S Y N G I
D S
T T
T P V
T S S
L H
G A
G M
S E
K G Y F G R S F W L A V A N T S S A T I S S A G S V F V V I A L L F
Microvesicle
membrane F S W G I F L F S L F L F Y A N I S L L L T A I E A A L L S T W V L V G R V F S I L L N F Y D C P L
V V
W F L G P I P F G Y E
E R K K
H S
E Q P E D A P L
L T
T N E P T-COOH P
Y K
P A A G R R
F H K S R F
G G Q A
Cytoplasm
V L Q N V V D M-NH2

26. Transmembrane structure of pantophysin
V N E E I F A S F N Y P F R L M
T S I V A L
S Q
P S
Intra-
vesicular
domains
P P
IV1
S D V C
Lumen of
microvesicle
G T P V Q Y
T R
K C
T D S
Q K N G
A G
V T T E S
W N
IV2
N F
K C T
I L
V A G
S Y N G I
D S
T T
T P V
T S S
L H
G A
G M
S E
K G Y F G R S F W L A V A N T S S A T I S S A G S V
Trans-
Membrane
(TM) domains F V V I A L L F
Microvesicle
membrane F S W G I F L F S L F L F Y A N I S L L L T A I E A A L L S T W V L V G R V F S I L L N F Y D C P L
V V
W F L G P I P F G Y
Cytoplasmic
(Cyt) domains E
E R K K
H S
E Q P E D A P L
L T
T N E P T-COOH P
Y K
P A A G R R
F H K S R F
G G Q A
Cytoplasm
V L Q N V V D M-NH2

27. Genealogy of PanI alleles in the Atlantic cod PanIA alleles (N = 64)
BA105A
Genealogy of PanI
alleles in the
Atlantic cod
BA108A
BA107A
BS39A
BA143A
BA112A
BS21A
BS29A
IC74A
IC70A
BS71A
BA115A
BA126A
BS72A
BA128A
BS49A
BS81A
BA132A
BS53A
BS87A
NS1A
NS12A
NS79A
NS73A
NS28A
PanIA alleles
(N = 64)
100
NS34A
NS91A
NS41A
NS74A
NS58A
IC30A
IC2A
NF42A
NF24A
NF142A
NF94A
NF88A
NF162A
NF158A
BA140A
BA138A
BA149A
BS20A
NF17A
NS83A
NF73A
BS31A
BS64A
NS70A
NS68A
IC8A
IC6A
IC9A
IC80A
IC41A
IC42A
IC78A
IC61A
NF6A
NF11A
NF56A
NF36A
BA105B
BA108B
BA107B
BA112B
NS1B
BA128B
BA126B
BA115B
NF88B
BA138B
BA132B
BA143B
BA140B
BA149B
BS20B
BS21B
BS31B
BS29B
BS39B
BS53B
BS49B
PanIB alleles
(N = 64)
BS64B
BS71B
BS81B
BS72B
BS87B
NS12B
NS68B
NS41B
NS28B
NS70B
NS91B
NS74B
IC6B
IC2B
IC8B
IC30B
IC25B
IC41B
100
IC42B
IC61B
IC70B
IC78B
IC74B
NF11B
IC80B
NF36B
NF24B
1 change
NF73B
NF56B
NS34B
NF158B
NF94B
NF17B
NF6B
NS58B
NF162B
NS73B
NS83B
NS79B
NF42B
Gadus ogac
NF142B

28. Amino acid differences between PanIA and PanIB alleles
V N E E I F A S F N Y P F R L M
T S I V A L
S Q
P S
P P
S D V C
Lumen of
microvesicle
G T P V Q Y
T R
K C
IV1
T D S
Q K N G
A G
V T T E S
W N
N F
K C T
I L
V A G
S Y N G I
D S
T T
T P V
T S S
L H
G A
G M
S E
K G Y F G R S F W L A V A N T S S A T I S S A G S V F V V I
Microvesicle
membrane A L L F F S W G I F L F S L F L F Y A N I S L L L T A I E A A L L S T W V L V G R V F S I L L N F Y D C P L
V V
W F L G P I P F G Y E
E R K K
H S
E Q P E D A P L
L T
T N E P T-COOH P
Y K
P A A G R R
F H K S R F
G G Q A
Cytoplasm
V L Q N V V D M-NH2

29. Amino acid substitutions within PanI allelic classes
___________________________________________________________
Codon Amino Acid Distribution
Allele Position Change Location Classificationa in sample
PanIA Lys to Gln IV Radical Fixed
Asn to Thr IV Radical Fixed
Ser to Thr IV Radical Fixed
PanIB Glu to Val IV Radical Fixed
Lys to Asn IV Radical Fixed
Asn to Asp IV Radical Fixed
a following Taylor (1986)