1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
|
<!-- header fragment for html documentation -->
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Final//EN">
<HTML>
<HEAD>
<META NAME="description" CONTENT="Estimation of population parameters using genetic data using a maximum likelihood approach with Metropolis-Hastings Monte Carlo Markov chain importance sampling">
<META NAME="keywords" CONTENT="MCMC, Markov chain, Monte Carlo, Metropolis-Hastings, population, parameters, migration rate, population size, recombination rate, maximum likelihood">
<TITLE>LAMARC Documentation: Bayesian Tutorial</title>
</HEAD>
<BODY BGCOLOR="#FFFFFF">
<!-- coalescent, coalescence, Metropolis-Hastings, Markov chain Monte Carlo
simulation, migration rate, effective population size, recombination rate,
maximum likelihood -->
(<A HREF="forces.html">Previous</A> | <A HREF="index.html">Contents</A> | <A HREF="search.html">Next</A>)
<H2>What do the population parameters mean?</H2>
<P> This page explains what each of the population parameters
estimated by LAMARC means, and what its units are. The same information
is available elsewhere in the documentation but it is collected here for
convenience.</P>
<UL>
<LI><A HREF="parameters.html#theta">Theta</A></LI>
<LI> <A HREF="parameters.html#m">M (migration rate)</a></LI>
<LI> <A HREF="parameters.html#e">Epoch boundary time </a></LI>
<LI> <A HREF="parameters.html#r">r (recombination rate)</a></LI>
<LI> <A HREF="parameters.html#g">g (exponential growth rate)</a></LI>
<LI> <A HREF="parameters.html#alpha">alpha (shape parameter of gamma distribution)</a></LI>
</UL>
<H3><A NAME="theta"> Theta </H3>
<P> Theta is two times the mutation rate (per site per generation) times the
number of heritable units in the population. Thus, if you are considering
diploid individuals (with two heritable units each) Theta is 4Nmu. If
you have haploids, it is 2Nmu. For mitochondrial DNA, which is heritable
only in females, it is Nmu (or 2N(f)mu).</P>
<P> If you wish to combine genes with different copy numbers, be sure to
set the relative N appropriately or you will get a confused result. For
example, if you are combining mitochondrial DNA and nuclear DNA, either
set the relative N as 1 for mtDNA and 4 for nuclear (you will then estimate
Theta on the mtDNA scale) or as 0.25 for mtDNA and 1 for nuclear (you will
then estimate Theta on the nuclear scale).</P>
<P> Similarly, if you are combining data with different mutation rates, be
sure to set the relative mu appropriately. Theta will be scaled by whichever
mutation rate you select to be 1. So if you believe your microsatellites
mutate 100 times faster than your DNA, either set the msat rate to 100 and
the DNA rate to 1 (you will then estimate Theta in the DNA scale) or
set the msat rate to 1 and the DNA rate to 0.01 (you will then estimate
Theta in the msat scale).</P>
<P> If you wish to convert Theta to a headcount of individuals, you will
need both an external estimate of the mutation rate, and an idea of
the ratio between headcount population size and effective (breeding)
population size. LAMARC cannot help you with either of these issues.</P>
<H3><A NAME="m"> M (migration rate) </H3>
<P> LAMARC measures migration rate as M=m/mu, where m is the chance
for a lineage to immigrate per generation, and mu is the mutation rate
per site per generation. An M of 1 means that it is as likely for
the sequence to migrate as it is for a site on the sequence to mutate.
If there are different mutation rates for different genes, M will be
given relative to the gene (if any) given a mutation rate of 1.0.<P>
<P> Biologists often want to measure migration rate as 4Nm.
This is useful because when 4Nm is higher than one,
the force of migration becomes strong enough to compete with
genetic drift.
To get 4Nm, multiply LAMARC's estimate of
M by its estimate of Theta for the <B> recipient</B> population. </P>
<P> LAMARC is the wrong tool to estimate migration rates if 4Nm
is larger than about 5, as it will go crazy trying to keep track of
so many migration events. If the program STRUCTURE sees no structure
in your populations, there is probably too much migration for LAMARC
to succeed, and you may need to pool your populations together.</P>
<H3><A NAME="e"> Epoch boundary time </H3>
<P> Epoch boundary times are estimated as number of generations
back from the present (the time at which the data were collected) to
the population splitting time, scaled by the mutation rate mu per
site per generation. Thus, an epoch boundary time of 1 means
that on average each site will have accumulated one mutation since
the populations split, and would represent a very ancient split. If
you want to know how many generations or years ago the populations
split, you will need external information about mu.</P>
<P> Note that not all scenarios will allow inference of epoch boundary
times, migration rates, and population sizes. For example, if the
populations split yesterday there will be no information about their
post-split sizes, and if they split very long ago there will be
no information about their pre-split sizes. If migration between them
is very high there will be no information about the split time. It
is important to compare the Bayesian posteriors with their priors; a
posterior that closely resembles its prior indicates a parameter about
which the data can say nothing.</P>
<H3><A NAME="r"> r (recombination rate) </H3>
<P> LAMARC measures recombination rate as r=C/mu, where C is the
rate of recombination per inter-site link per generation, and mu
is the mutation rate per site per generation. An r of 1 means that
it is as likely for a recombination to occur next to a site as it
is for that site to mutate, and is a rather high rate of recombination.
If there are different mutation rates for different genes, r will be
given relative to the gene (if any) given a mutation rate of 1.0.</P>
<P> LAMARC does not allow r to vary among populations.</P>
<P> For many comparative purposes you will want the recombination rate
per locus rather than per site; you can obtain this by multiplying
LAMARC's r by the number of sites minus 1 (as no meaningful recombination
can occur rightward of the final site).<P>
<H3><A NAME="g"> g (exponential growth rate)</H3>
<P> This is the hard one!</P>
<P> The parameter g is the exponent of an exponential growth rate
formula which gives the Theta at a time t in terms of the modern-day
Theta and the growth rate:</P>
<P> Theta(t) = Theta(modern) exp(-gt) </P>
<P> Time is measured in mutational units; that is, one unit of time is the
time needed for each site to experience, on average, one mutation.
The units of our mutation rate are mutations per generations, so the
units of g are 1/generations. Almost no one finds this intuitive.</P>
<P> We define time as increasing into the past (the present is time 0)
and as a result, a negative value of g indicates that the population has
been shrinking (it was bigger in the past than it is now) and a positive
value indicates that it has been growing (it was smaller in the past than
it is now). A g of zero indicates constant size. This much we know
even without knowing the mutation rate, but to say anything more we
need to either know the mutation rate, or be comparing two populations
with the same mutation rate (in which case, the one with higher g is
growing faster). If we know the mutation rate, we can plug it into the
equation above to get a feeling for what this implies in terms of
Theta or effective population size.</P>
<P> Be aware that LAMARC's estimates of g are biased upwards (due to
the shape of the likelihood surface) especially with only one or a few
genes. If the estimate of g is positive and big, but the confidence
intervals include zero, it's quite likely that there is in fact little or no
growth. If the intervals exclude zero, that finding is generally
reliable.</P>
<H3><A NAME="alpha"> alpha (shape parameter of gamma distribution) </H3>
<P> LAMARC can use the gamma distribution to represent the unknown
variation of mutation rate among genes. Gamma was chosen because it is
a simple distribution that ranges from looking rather like an exponential
(most genes mutate very little, a few mutate much more rapidly) to
looking rather like a bell curve (there is an average mutation rate and
genes are spread nearly symmetrically around it).<P>
<P> The gamma distribution has two parameters, but we fix the mean of
the gamma to 1, which determines one parameter. This leaves only the shape parameter
alpha (α). Low values of alpha, below 1, describe a gamma distribution
which looks somewhat exponential. High values describe a gamma which looks somewhat like
a normal. So if you estimate alpha as 0.3, this means that most of your
genes have a low mutation rate but a few are much higher; if you estimate
gamma as 20, this means that your genes are tightly clustered around
their mean mutation rate. If all your regions mutate at exactly the same
rate, the proper estimate of alpha would be infinite, but LAMARC will
estimate an arbitrary high value that is practically equivalent to infinity.</P>
(<A HREF="forces.html">Previous</A> | <A HREF="index.html">Contents</A> | <A HREF="search.html">Next</A>)
<!--
//$Id: parameters.html,v 1.7 2012/05/16 17:14:01 mkkuhner Exp $
-->
</BODY>
</HTML>
|
