Transcript 4.4 II
MAT 4830
Mathematical Modeling
4.4
Matrix Models of Base
Substitutions II
http://myhome.spu.edu/lauw
Markov Models
Review of Eigenvalues and Eigenvectors
An example a Markov model.
Specific Markov models for base
substitution:
• Jukes-Cantor Model
• Kimura Models (Read)
Recall
Characteristic polynomial of A
Eigenvalues of A
Eigenvectors of A
Recall
Characteristic polynomial of A
P( ) det( A I )
Eigenvalues of A
zeros of P( )
Eigenvectors of A
( A I ) x 0, x 0
Geometric Meaning
Consider A : R R
n
n
( A I )x 0
Ax x
scalar multiple of x
Lemma
A x x , for n Z
n
n
Recall
Use the transition matrix, we can
estimate the base distribution vectors pk
of descendent sequences Sk , k 1, 2,3,...
by
pk Mpk 1
An example of Markov model
Markov Models Assumption
What happens to the system over a given
time step depends only on the state of the
system and the transition matrix
Markov Models Assumption
What happens to the system over a given
time step depends only on the state of the
system and the transition matrix
In our case, pk Mpk 1
pk only depends on pk-1 and M
Markov Models Assumption
What happens to the system over a given
time step depends only on the state of the
system and the transition matrix
In our case, pk Mpk 1
pk only depends on pk-1 and M
Mathematically, it implies
P Sk sk | Sk 1 sk 1 Sk 2 sk 2
P Sk sk | Sk 1 sk 1
S0 s0
Markov Matrix
All entries are non-negative.
column sum = 1.
PA| A
P
G| A
M Pi| j
PC| A
PT | A
PA|G
PA|C
PG|G
PG|C
PC|G
PC|C
PT |G
PT |C
PA|T
PG|T
PC|T
PT |T
Markov Matrix : Theorems
Read the two theorems on p.142
Jukes-Cantor Model
Jukes-Cantor Model
Additional Assumptions
• All bases occurs with equal prob. in S0.
1
p0
4
1
4
1
4
1
4
T
Jukes-Cantor Model
Additional Assumptions
• Base substitutions from one to another are
equally likely.
PA| A
P
G| A
M Pi| j
PC| A
PT | A
Pi| j constant
3
PA|G
PA|C
PG|G
PC|G
PG|C
PC|C
PT |G
PT |C
, for i j
PA|T
PG|T
PC |T
PT |T
Jukes-Cantor Model
PA| A
P
G| A
M Pi| j
PC| A
PT | A
Pi| j constant
3
PA|G
PA|C
PG|G
PC|G
PG|C
PC|C
PT |G
PT |C
, for i j
PA|T
PG|T
PC |T
PT |T
Jukes-Cantor Model
1
/ 3
M Pi| j
/ 3
/ 3
Pi| j constant
3
/3
1
/3
/3
/3
/3
1
/3
, for i j
/ 3
/ 3
/ 3
1
Observation #1
1 prob. of no base sub. in a site for 1 time step
prob. of having base sub. in a site for 1 time step
rate of base sub. sub. per site per time step
Mutation Rate
Mutation rates are difficult to find.
Mutation rate may not be constant.
If constant, there is said to be a
molecular clock
More formally, a molecular clock
hypothesis states that mutations occur at
a constant rate throughout the
evolutionary tree.
Observation #2
1
/ 3
M Pi| j
/ 3
/ 3
/3
1
/3
/3
1
p0
4
T
1
4
1
4
1
4
p1 ?
pk ? for k 1, 2,3,...
/3
/3
1
/3
/ 3
/ 3
/ 3
1
Observation #2
1
/ 3
Mp0
/ 3
/ 3
/3
1
/3
/3
/3
/3
1
/3
p1 ?
pk ? for k 1, 2,3,...
1
4
/ 3 1
/ 3 4
?
/ 3 1
1 4
1
4
Observation #2
The proportion of the bases stay
constant (equilibrium)
What is the relation between p0 and M?
Example 1
What proportion of the sites will have A
in the ancestral sequence and a T in the
descendent one time step later?
1
/ 3
M Pi| j
/ 3
/ 3
/3
1
/3
/3
/3
/3
1
/3
/ 3
/ 3
/ 3
1
1
p0
4
1
4
1
4
1
4
T
Example 2
What is the prob. that a base A in the
ancestral seq. will have mutated to
become a base T in the descendent seq.
100 time steps later?
Example 2
What is the prob. that a base A in the
ancestral seq. will have mutated to
become a base T in the descendent seq.
100 time steps later?
pk Mpk 1
p100 M 100 p0
Example 2
p100 M 100 p0
p100
M 100
p0
Example 2
p100 M 100 p0
p100
M 100
p0
100
M4,1
"Must be" P(S100 T | S0 A)
Example 2
p100 M 100 p0
p100
M 100
p0
100
M4,1
"Must be" P(S100 T | S0 A)
[]
If M 100 P ( S100 i | S0 j )
then p100 M 100 p0
[ ]
If p100 M 100 p0
then M 100 P ( S100 i | S 0 j )
Example 2
p100 M 100 p0
p100
[]
M 100
p0
100
M4,1
"Must be" P(S100 T | S0 A)
If M 100 P ( S100 i | S0 j )
then p100 M 100 p0
[ ]
If p100 M 100 p0
then M 100 P ( S100 i | S 0 j )
For general n, can be prove by inductive arguments.
Homework Problem 1
p2 M 2 p0
p2
M2
p0
2
Explain why M4,1
P(S2 T | S0 A)
Example 2 (Book’s Solutions)
pt M t p0
pt
Mt
p0
P(St T | S0 A) ?
Find the eigenvalues i and the
corresponding eigenvectors vi ,
for i 1, 2,3, 4
Example 2 (Book’s Solutions)
pt M t p0
pt
1
t 0
M
0
0
Find the eigenvalues i and the
Mt
p0
corresponding eigenvectors vi ,
for i 1, 2,3, 4
Example 2 (Book’s Solutions)
pt M t p0
pt
1
t 0
M
0
0
Find the eigenvalues i and the
Mt
p0
corresponding eigenvectors vi ,
for i 1, 2,3, 4
1
0
1v 1v 1v 1v
0 4 1 4 2 4 3 4 4
0
Example 2 (Book’s Solutions)
pt M t p0
pt
1
t 0
M
0
0
Find the eigenvalues i and the
Mt
p0
corresponding eigenvectors vi ,
for i 1, 2,3, 4
1
0
1v 1v 1v 1v
0 4 1 4 2 4 3 4 4
0
Example 2 (Book’s Solutions)
pt M t p0
pt
1
t 0
M
0
0
Mt
p0
1
1
1
1
1
t 0
M t v1 M t v2 M t v3 M t v4
M
0 4
4
4
4
0
1
1
1
1
1t v1 2t v2 3t v3 4t v4
4
4
4
4
1 3 3 t
1
4 4 4
t
1 1
3
1
4 4 4
t
3
1
1
1
4 4 4
t
1 1 3
1
4 4 4
Example 2 (Book’s Solutions)
1 3 3 t
1
4 4 4
t
1 1
3
1
4 4 4
Mt
t
1 1 1 3
4 4 4
1 1 3 t
1
4 4 4
1 1 3
1
4 4 4
t
1 1 3
1
4 4 4
t
1 3 3
1
4 4 4
t
1 1 3
1
4 4 4
t
1 1 3
1
4 4 4
t
1 3 3
1
4 4 4
t
1 1 3
1
4 4 4
t
1 1 3
1
4 4 4
t
t
1 1 3
1
4 4 4
t
1 1 3
1
4 4 4
t
1 1 3
1
4 4 4
t
1 3 3
1
4 4 4
Our Solutions
Theorem:
Suppose M is a symmetric matrix with eigenvalues i and the
corresponding eigenvectors vi .
Let P [v1 v1
1
2
vn ] and D
0
Then, D P 1MP
0
n
Our Solutions
1t
0
t
M P
0
0
0
2t
0
0
0
0
3t
0
0
0 1
P
0
t
4
D P1MP
Our Solutions
1 3 4 t
1
4 4 3
t
1 1
4
1
4 4 3
Mt
t
1 1 1 4
4 4 3
1 1 4 t
1
4 4 3
1 1 4
1
4 4 3
t
1 1 4
1
4 4 3
t
1 3 4
1
4 4 3
t
1 1 4
1
4 4 3
t
1 1 4
1
4 4 3
t
1 3 4
1
4 4 3
t
1 1 4
1
4 4 3
t
1 1 4
1
4 4 3
t
t
1 1 4
1
4 4 3
t
1 1 4
1
4 4 3
t
1 1 4
1
4 4 3
t
1 3 4
1
4 4 3
Maple: Vectors
Maple: Vectors
Homework Problem 2
Although the Jukes-Cantor model
T
p
0.25
0.25
0.25
0.25
,a
assumes 0
Jukes-Cantor transition matrix could
describe mutations even a different p0 .
Write a Maple program to investigate the
behavior of pk .
Homework Problem 2
Homework Problem 3
Read and understand the Kimura 2parameters model.
Read the Maple Help to learn how to find
eigenvalues and eigenvectors.
Suppose M is the transition matrix
corresponding to the Kimura 2-parameters
model. Find a formula for Mt by doing
experiments with Maple. Explain carefully your
methodology and give evidences.
Next
Download HW from course website
Read 4.5