A Model of Evolution
A Model of Evolution
A Model of Evolution
M. Bunge
Foundations and Philosophy of Science Unit, McGill University, Montreal H3A1W7,
Canada
(Received March 1978; revised May 1978)
0307-904X/78/0203-0201 S02.00
©1978 IPC Business Press Appl. Math. Modelling, 1978, Vol 2, September 201
A m o d e l o f evolution: M . B u n g e
The average number of defectors from variety V/ simultaneously at the next generation and the
between times t and t + 1 will be: mutation-and-selection process proceeds relentlessly.
Every one of the n varieties prospers once it has
D ~ = 2 ~ m o o -- di)N~ (3)
emerged, and the prosperity of each is conducive to
that of all. (The only qualitative change in this process
A n d the new arrivals to variety V~ in the same time can be induced by a sudden environmental change
interval number: causing some death rates to jump to unity. ) But, in
A ~ = 2 ~ mj,(l -- d~)Ni (4) general, the process will not be 'explosive':while some
entries of the evolution matrix may be positive, others
may be zero or negative. An overall negative matrix
for they will be drawn from all varieties but Vi. The corresponds of course to extinction.
factor 2 in D ~ and A ~ shows that we are making the As long as the death rates di and the mutation
additional hypothesis that mutants come in pairs--i.e. probabilities mu are not specified, the model is a
if an organism is a m u t a n t so is its sister. phenomenological skeleton.
Collecting equations (1), (3) and (4) we obtain the
net average population of variety V~ at time t + 1:
Underlying mechanism
.N[~+I = 2mi,~l -- di)N[ - D i + A i
Let us now specify the architecture of our evolving
organisms as well as their mutability (mu) and viability
= 2 [ m i , l - di) - ~ ~ j { 1 - di)]N[
(d~). We assume that the n varieties of our
biopopulation differ by the composition or
+ 2 ~ ,nj,{1 - - d j ) N l (5) arrangement of a fixed number of modules or basic
units that can be repeated. For the sake of simplicity
we suppose that there are just two kinds of module, to
Because of the normalization condition (2) of the be called 0 and 1. Thus, if each entity is composed of
mutation probabilities, the total probability of a four such modules arranged differently, we obtain a
mutation from V~is: total of n = 24 = 16 varieties, namely 0000, 1000,
n~j = ~ rno -- ma = 1 - mu (6) 0100, 0010, 0001, 1100, 0110, 0011, 1010, 0101, 1001,
j~i j 1110, 0111, 1011, 1101, and 1111. The general formula
for a variety whose members are composed of m
Introducing this value into equation (5) we are finally modules is abc.., m, where each letter is a binary digit.
left with: So much for architecture.
N~+ ~ = 2(2mu - 1)(1 -- d,)N~ Every change in at least one of the m binary digits
occurring in the formula for a variety is a mutation.
+ 2 ~ mj,O -- d~)Ni (7) Thus a mutant will differ from its parent organism
either by the total number of O's (or l's) or by the
This system of n linear difference equations order of O's and l's. Thus in the above example the
summarizes our model of evolution. The solution to transitions 0000 ~ 1000, 0100 ~ 0110, and 0111 ~ 1111
these equations can be obtained as follows. Set: are mutations and, at the same time, changes in
/v; variety. Of course not all mutations are equally
probable. We shall determine certain relations among
mutation probabilities with the help of the following
considerations on viability.
a =-- i1%tl -- II2(2m. - 1)(1 - d3~oll N~
Considering the environment, we lump all the
b = ltbi~ll = 112mall -- d~)(1 -- 6~j)lt N, = - (8) environmental factors into a single, real valued
function e with values in the [ - 1, 1] interval. And we
assume that the probability that an organism
N~ belonging to variety V~survives up to and including
division is symmetrical about the origin. More
where f o = 1 if i = j and vanishes otherwise. The precisely, we assume that the death rates di in equation
system, (7), can now be condensed into the matrix (1), far from being constant, depend upon the
difference equation: environmental variable e in the following way:
N l ÷ 1 = (a + br)N~ (9)
~ri(l -- e 2) for -- 1 ~< e ~ 1
where T designates the transposition operation. 1 -- di = ~1 otherwise (12)
Iterating equation (9) we find the general equation:
Nt+k = (a + br)kNt where ke N (10) where 0 ~< vi ~ 1 for each variety V/.
Since vi is characteristic of each variety and is
The solution to this equation is: environment-independent, we call it the "adaptedness"
Nt = (a + br)tNo (11) of organisms of the variety I'j. (Notice the difference
between this concept and that of fertility. Adaptedness
where No is the column matrix of the initial is a necessary condition for fertility, so the latter is an
populations of the n varieties being considered. indicator of the former.)
Suppose that all the entries of the evolution matrix The factor (1 - e 2) represents the-selection effected
a + b "r are positive and that initially a single variety is by the environment. A substantial increase or decrease
present. Then all the other varieties can emerge in e will have a considerable impact on the fraction
B
~1 -- m2z -- nO m mz2 141 -- mzz -- nO
vl=v,=v f o r 0 0 a n d 11 (13) (1 - nqt)[I --mzz--ml, 1 --m1~ ! --nqt
II 3--mzz - m 3-mzz--m 33-- mzz-- m m t t
Va = v3 = cry, with e > 1, for 01 and 10
We now proceed to trace the evolution of our four-
Finally we relate viability to mutability, namely: we
variety biopopulation. To compute their numbers at
postulate that a mutation is the more probable the less
successive generations we must k n o w their initial
viable its product is. More precisely, we assume:
numbers. Suppose a single .variety is initially present,
m U >1 mik if and only if vj <~ r'k (14) say 1/i, or 00. Call N~ = N its initial population. Then
by equation (7), one generation later, the average
We can now build the matrix of the mutation numbers are:
probabilities in the simple case where rn = 2, whence
n = 4. In this case the viabilities are those indicated in N~ = 2 ( 2 m x x - 1)(1 -- d , ) N
equation (13). Hence the chances of reproducing are:
N2= 2(1--mtl) (1--dl)N
1 - d I = l -- d . = v(1 -- e 2) 3 - - m 2 2 - - #z
(15)
1 - d 2 = 1 - d 3 = :t.t~l -- e 2) (24)
= N,
The hypothesis, equation (14)~ that most mutations are
deleterious, together with the assumption, equation (13), N~ = (l - "22 - m)N~
about the adaptedness values, entails:
If ml 1 ~< ½ or d, = I, the first variety becomes
m41 = m14 ?na2 = m23 = m (16) extinct forthwith. A n d if ml, = 1 while d t < I, Vt
(where this m is not to be confused with the number of prospers without evolving. The necessary and sufficient
kinds of modules per organism, which we have taken condition for the evolutionary process to occur is then:
to be 2), and:
½<m11<1 and 0~<dI<l (25)
m21 = film12 real = fllmaa
(17) If this condition obtains then the lower bound of the
m24 .= f12m42 m 3 4 ----- f 1 2 m 4 3 , population of varieties 1/2 and V3 at first generation is:
with ill, fiE > 1. We-simplify by setting fll = f12 = fl :](1 -- roll)(1 -- da)N
and so are left with:
Since this number equals at least 2, we infer that the
m2l = tirol2 m31 = flmxa
(18) lower b o u n d of the stability of the parent variety is
m24 = tim42 m34 = flm43 given by:
mll I> 3/(1 - dx)N (26)
Simplifying assumptions a n d final e q u a t i o n s Because the upper bound of m , , is I, we infer that for
Since permutations of modules are less radical than a change of variety to occur it is necessary that:
changes in composition, we can make the further
N > 3/1 -- dt (27)
simplifying assumptions that:
For a low death rate of the parent variety, as few as 4
m12 = m13 m24 = m34 (19)
members of it could give rise to the differentiation
Similarly, we posit that the variety O1 is just as stable process.
as 10, and that' O0 and l 1 are equally stable: The above conditions concern only varieties V2 and
V3. For variety V4 to emerge a more severe condition
m22 = ma3 nh, = m44 (20)
must be met, as we see from the last of equations (24),
Our final assumption is that the mutations between the namely, equation (25) plus (m22 + m < I) or,
extreme varieties O0 and 11 proceed via the equivalently, equation (25) plus (m2 x + m24 > 0).
intermediate species O1 and lO, i.e.: Finally we write the general evolutionary equations
(7) for an arbitrary t value. Using the mutation matrix,
II114 = IH1211124 + FH13tH34.
(21) equation (23), we obtain the following System of
m 4 1 = D142Yl121 "~- !Y14317131 first-order linear difference equations:
I~li~IOTHEEK MATHEMATISCH tiff.N~F, UM
L . AMSTiRDAM
Appl. Math. Modelling. 1 9 7 8 . V o l 2. S e p t e m b e r 203
A model of evolution: M. Bungo
N:+a = 2(2mal -- 1)(1 -- dl)N: evolution, equations (7), could be turned, from a mere
descriptive black box, into a mechanismic model
+ (I -- rn22 -- m)(1 -- d2)N2
representing the mutation-selection process.
+ (I -- m22 -- m)(l -- d3)N~
+ 2(1 - m~)(1 - m ~e - m)(1 _ d,)N,"
Discussion
3 -- m22 -- m
O u r model of evolutionary processes is perhaps the
N ? + x = 2(2m22 - 1)(1 - d2)N~ simplest that one can imagine, as it deals with
2(1 -- m11) idealized bacteria assumed to be composed of two
+ (1 - a~)N: units of the same or different kinds. It is a linear
3 -- m22 ~ m
model. Furthermore, it includes a n u m b e r of
simplifying assumptions. A n d yet the final system of
+ 2m(l -- ds)N~ + 2 3 (1 - m11) (1 - d4)N~
m22 m m equations is sufficiently general to invite the use of a
digital computer. This suggests that a more realistic
N~. 1 = 2(2m33 -- 1)(1 - d3)N~ model, say of a plant or animal lineage, must be far
more complicated. However, this should be taken as a
2(1 - roll ) challenge rather than a pretext for refraining from
-(1- d~)~
+ 3 -- m22 -- m modelling evolutionary processes.
The use of our model is presumably methodological
2(1 -- rn 11) rather than substantive. T h a t is, it shows the kind of
+ 2m(1 -- d2)N~2 + .(1 -- d,)N~
~ m22 -- m premises (ddta and specific hypotheses) that must be
added to the extremely general theory of evolution in
N:+z = 2(2m44 -- 1)(1 -- d.,)N:
order to build a precise, albeit highly idealized, model
+ 2(1 -- mix)(1 -- m22 -- m)(l of an evolutionary process. O u r model also shows that
d l)Ntl it is not biospecies but biopopulations (or even entire
3 - m22 - m
communities) that do the evolving. (Species are sets,
+ fl - m 22 - m)(1 - d2)N~ and sets are concepts, not objects that are capable of
changing.) Finally, our model shows that the theory of
+ (1 - - m2z -- m)(1 -- d3)N] evolution, far from-being nonscientific, is a rich matrix
W e shall not a t t e m p t to solve this system. O u r aim for the gestation of any n u m b e r of specific, hence
was only to show how the general equations of precisely testable, scientific theories.