A PROBABILISTIC THEORY OF WITHIN SPECIES EVOLUTION David J. Murrow July 8, 2001 A Probabilistic Theory of Within Species Evolution—D.J. Murrow The evolution of a species over time is modelled in this paper as a random process which is governed by the laws of probability. That this is a realistic and practical model of the actual process of evolution is left to anthropologists, philosophers, and clergymen. What is shown herein is merely that, if a species is affected by life events which follow the laws of probability similar to those modelled herein, then evolution of the species will occur over time. The key features of the theory modelled herein are: 1) Survival/Life Expectancy 2) Natural Selection 3) Procreation Proclivity 4) Inheritance Consider that a member of a certain species of animal life posseses a set of characteristics , x , which can be quantified as an Nx1 vector, where each component of the vector is a single characteristic measured by a real random number. For the purposes of this paper, if the statistics (mean, variance, pdf, etc ) of this random variable vary over time from generation to generation, the species is said to be evolving. Survival/Life Expectancy Each member k of the species is considered to have a life duration in time, which depends in a probabilistic, or random, way on the (survival) characteristics xk possesed by the member. Dependent random variables have a joint pdf which is the product of the marginal pdfs with the associated conditional pdf, i.e. p( k , xk ) p / x ( k / xk ) p x ( xk ) p x / ( xk / k ) p ( k ) . The marginal pdfs are defined by p x ( x) p( , x)d ; p ( ) p( , x)dx and the conditional pdfs are defined by p / x ( k / xk ) p( k , xk ) / p x ( xk ); p x / ( xk / k ) p( k , xk ) / p ( k ) The expected, or mean, value of the life duration of an individual member of the species is termed the life expectancy of the species and is given by 2 E k k p ( k )d k k p( k , x k )d k dxk k p( k / xk ) p x ( xk )d k dxk E k / xk p x ( x k )dxk / x ( xk ) p x ( x k )dxk where p(k) is the marginal pdf associated with the life duration of the species member and px(xk) is the (multi-dimensional) marginal pdf associated with the physical characteristics of the species member. The function x) is termed the conditional lifespan of the species and is given by / x ( x) k p / x ( k / x)d k Several things should be noted at this juncture. First, the life duration of a single species member is a single real (random) number. Therefore, the first integral above is a single integral. Second, since the physical characteristics xk of a species member is an N x 1 vector , the associated pdf is multi-dimensional and the second integral above is multidimensional. Finally, the physical characteristics of an individual member of the species is assumed not to vary with time over its lifetime. This simplifying assumption is a conservative one with respect to species evolution in that usually the positive physical characteristics of a particular animal increase from childhood to adulthood. By neglecting this time dependence of the characteristics over the lifetime of a single individual, if we still find that a long term generational time dependence in the statistics of the characteristics can be shown, this more strongly shows that the species is evolving. The nature of the species conditional lifespan function /x(x) will be shown to strongly influence the evolution of a species. Consider the variation of a single component of the physical characteristics xn , with the other components held fixed in value, as shown in Figure DJM-1. /x(x) D C A B xn Figure DJM-1 Species Conditional Lifespan vs Physical characteristic Function Types 3 In case A above, the species conditional lifespan function increases monotonically with its nth physical characteristic. In other words, the greater the value of the measure of the nth characteristic xn , the greater the lifespan of the a species member with the given physical characteristic value . In case B above, the species conditional lifespan function decreases monotonically with its nth physical characteristic. In other words, the greater the value of the measure of the nth characteristic xn , the lesser the lifespan of the a species member with the given physical characteristic value . In case C above, the species conditional lifespan function decreases with its nth physical characteristic until it reaches a minimum and then increases for greater values of the nth physical characteristic. In case D above, the species conditional lifespan function increases with its nth physical characteristic until it reaches a maximum and then decreases for greater values of the nth physical characteristic. Survival/life expectancy dependence on physical characteristics alone will not produce evolution within a species. Procreation and inheritance are also required. Natural Selection Not all species mate for procreation. Those that do are divided into two sexes, male and female. Natural selection refers to the tendency in prospective mates from the two sexes to prefer values of x in their partner's physical characteristics which enhance the survival/life expectancy of the species. Let member #m be a male member of the species and member #f be a female member. Then xm and xf are the two sets of physical characteristics possesed by the two members. The corresponding conditional life expectancies are x(xm) and x(xf). A reasonable model for natural selection would call for the likelihood of mating between two members of the opposite sex with drastically different conditional life expectancies to be small. Let m,f = xm , xf ) be the average rate of mating between male member m and female member f, and let m,f = x(xm) x(xf) be the difference in their conditional life expectancies . A reasonable model for natural selection would appear as in Figure DJM-2. Even if the life expectancies of two members are close, they may not mate due to failure to meet or other factors. Let pM(m,f) be the probability that two specified members of the opposite sex meet during their lifetime 4 m,f m,f Figure DJM-2 Natural Selection--Relationship Between Mating Rate and Conditional Life Expectancy let ym,f =1 if members m and f meet during their lifetime and ym,f =0 otherwise. Then a reasonable model for couples meeting is the binomial pdf py(m,f)=(pM(m,f))y (1-pM(m,f))(1-y) Procreation Proclivity In single sex species, the number of offspring produced by a member during its lifespan may be dependent on the physical characteristics xk of the member and on its lifespan k . A reasonable model for such procreation is the Poisson process, i.e., p nk (nk / x k , k ) ( k ) nk e k / nk ! nk=0,1,2,… where k = (xk)k In dual sex species, the number of offspring produced by a pair of opposite sex members during their lifespan may be dependent on the physical characteristics xm and xf of the members and on their lifespan m,f = min(m , f ) . A reasonable model for such procreation is the Poisson process, i.e., pn (nm, f / xm , x f , m, f , y m, f ) ( m, f ) nm , f e m , f / nm , f ! where m,f = (xm,xf)y(m,f)m,f . 5 Inheritance In single sex species, physical characteristics are passed on to a members offspring in a probabilistic way. Let the parent be member k, with physical characteristic xk and let the child be member j, with physical characteristic xj . The inheritance characteristic would be modelled by the conditional inheritance pdf pI(xj/xk). In dual sex species, physical characteristics are also passed on to pair of opposite sex members offspring in a probabilistic way. Let the parents be members m and f, with physical characteristic xm and xf and let the child be member j, with physical characteristic xj . The inheritance characteristic would be modelled by the conditional inheritance pdf pI(xj/xm , xf). Note that deterministic (non-random) inheritance can be forced by taking the pdfs to be Kronecker delta functions, e.g, p I ( x j / xk ) ( x j xk ) or p I ( x j / x m , x f ) ( x j ( x m x f ) / 2) In these cases the characteristics of the parent(s) would be passed on to the children in a fixed, deterministic manner. 6 Some example numerical results are next presented. The single sex species case is simpler mathematically and so is used in the following example. Consider a species of microbe whose dominant physical characteristic is , say, its length. At time zero, we may start with a small set (say 10) of members of the species whose lengths are random draws from a normally distributed pdf with a mean of 10 micometers and a standard deviation of 1 micrometer. This pdf is illustrated in Figure DJM-1a. Example: Microbe Size Evolution Initial Size Distribution Conditional Life Expectancy T/x=100/(1+((x-xT)/ =100/(1+((x-xT)/)4) p(x)=exp(-(x-xb)2/2 /22 )/(2 )/(22)1/2 xb=10 =1 Lifespan Distribution xT=20 =5 Procreation Proclivity conditional life expectancy =...50,60,70,80,... p(T)=4exp(-0.8( t/ t/))5 )( t/ t/))4 / / Figure DJM-1 Probabilistic Models for Example Microbe Species Characteristics From a survivability viewpoint, it is conceivable that a member of such a microbe species might be less likely to survive if it were too short or too long, and more likely to survive if it were near some optimal survival length. Such a function is shown in Figure DJM-1b, where an example conditional life expectancy function is shown plotted against the length of a microbe. This function peaks at 100days of life expectancy for a length in the neighborhood of 20micometers, and falls off rapidly on either side of the optimum survival length. The actual lifespan of a member is a random draw from some pdf with life expectancy as a parameter. An example pdf with reasonable characteristics is shown in Figure DJM-1c. A member microbe may produce a number of offspring during its lifespan. It is reasonable to take the expected rate of offspring production to be a fixed property of the microbe species. For example, suppose it is an offspring per 5 days on average, or 0.2 offspring per day. The expected number of offspring in a particular 7 members whose lifespan was 50 days would then be 10. A commonly used model for the actual number of offspring would be the Poisson pdf. The Poisson pdf for various mean values is illustrated in Figure DJM-2. Poisson Procreation Distribution Expected (mean) # Offspring=1 Expected (mean) # Offspring=10 Figure DJM-2 Poisson PDF with Various Expected Values For this example, the length of an offpsring microbe is taken here to be a normally distributed random variable with an expected (or mean) length equal to that of it's parent's actual length and a standard deviation of 1 micrometer. 8 Figure DJM-3 shows a single representative Monte Carlo trial of the resultant (random) population of the species versus time, along with the evolution of the species in length, lifespan, and number of offspring. This figure clearly shows that the microbe population is growing and that the species is evolving toward greater length, greater lifespan, and more offspring. Figure DJM-3 Representative Evolution of the Example Microbe Species Thus, under favorable probabilistic survival and propagation conditions, the average physical characteristics of a species will evolve over time. 9