======================================================================
Author: Vladimir F. Levchenko (Lew@ief.spb.su)
 Title: The Experience of Simulation of Evolutionary Process
======================================================================

       THE EXPERIENCE OF SIMULATION OF EVOLUTIONARY PROCESS

                       Vladimir F.Levchenko

      Institute of Evolutionary Physiology and Biochemistry
       of Russian Acad. Sci., St.Petersburg, 194223, Russia
                     E-mail: Lew@ief.spb.su


                             ABSTRACT

     The simulation model described below has been  developed  to
investigate the patterns and dynamics involved in the building of
evolutionary trees. The model is based upon darwinian theory.
     The  model  can  work  with  12   types   of   environmental
conditions. Up to 9 populations can  exist  in  any  environment.
Every population is described by 25 features. Mutations  are  the
changes of  features.  The  simulation  program  allows  to  draw
evolutionary trees on the screen of computer  and  to  write  the
results in a fail.
     Fortran version of program has  been  created  in  1987  for
PDP-11 and compatible computers and can be adapted for IBM PC.


                           INTRODUCTION

     The simulation model described below has been  developed  to
investigate the patterns and dynamics involved in the building of
evolutionary trees.  The  general  model  employed  assumes  only
certain specified ecological properties of  populations  and  the
basic principle of interactions  between  them,  and,  hence,  is
based upon darwinian theory.
     While the ultimate objective of this simulation  is  greater
insight into the patterns produced by natural  selection  working
overlong periods of time, any  simulation  of  evolution  remains
constrained by inherent problems arising from the large number of
calculations involved and the resultant effects of small  changes
in values at one time producing large changes in the system at  a
later point in time.
     It is important  to  note  biological  terminology  used  to
describe the work of model is a few metaphorical one. This is the
consequence of only some peculiarities of reality are enclosed in
model.
     The computer program was first developed in Fortran using  a
RT11 system in 1987.
     The program works by  cycle.  The  operations  on  the  each
cycles are related until 13 point of description of the model.

                       DESCRIPTION OF MODEL

     We can not simulate all, it is necessary to predetermine the
properties and features of "model  world"  and  the  features  of
object of the model.
     We have:
     1. 12 versions of "Environments".
     2. Each  environment  is  has  up  to  9  "Sub-environments"
(regions) each of which may be inhabited by only one  population.
These 9 populations may be drawn from one or more  species.  Some
sub-environments can be free.
     2.1.  In  the  model  we  don't  consider  the  quantity  of
specimens of population and treat each population as  having  the
same number of individuals.  More  exactly:  in  environment  can
exist doubles of  some  population.  Each  double  we  take  into
account as independent population.
     3. Any population is described 25 different  "Features".  In
the model all possible features are predetermined.  Total  number
of all possible features is 100 but description of any population
is some combination 25 features taken from this 100. For example:
dimension from 1M to  2M,  4  fins,  predatory,  roe...etc.  (See
below, moreover).
     4. We call as "Property" the group of similar features.  For
example: dimension (features here are: from 0.5 to 1, from  1  to
2,  from  2  to  5,...  for  instance),  extremities,   type   of
nourishment,.. etc. The total number of properties  is  25  also.
Each population have one feature from each property.
     5.  "Mutations"   is   change   of   any   feature(s)   into
property(ies) in the model. Special tables exist in the model  to
determine possible mutations into each property. For  example:  A
<-> B <-> C, where A,B,C are features  into  property.  Note:  to
transit from A to C for instance, it is necessary two step: A  ->
B -> C. We can change the frequency  of  mutation  on  population
(i.e. mutation/population) by means special parameter of model.
     6. The features, which can not exist in common we call  "Non
compatible features". For example: skull does  not  exist,  tooth
exist. Mutants which have as result of mutations such  prohibited
combinations of features are abolished by using special table  of
non compatible features.
     7. Moreover. Each feature have value from 0 to 10.  This  is
adaptive value. For instance: the value of fins into water =  10,
but on the land = 0. The special matrix exist in the model and in
this matrix values of all features for all  environment  are  set
using  biological  considerations.  This  are  subjective  values
certainly but it does not  influence  on  results  of  simulation
which are interesting for us (for instance, the dynamic  of  grow
of evolutionary trees).
     8. The "Survival" of population = summa of all  (25)  values
of features of population. Moreover, the survival = 0 if any  one
(two...) value(s) = 0. (We suppose in this case  that  population
have feature(s) non compatible with environment).
     We calculate in the cycle of work of  program  survivals  of
each population for each  environment,  although  any  population
exist only into one environment in fact.  This  is  necessary  to
decide the question about a transition of population on the  each
cycle. The survival of population may change after mutation.
     9. The  transition  of  population  from  initial  (for  it)
environment to other environment may take place  if  survival  of
this population is more in other environment. But for realization
of transition it is necessary:
     a) new environment has free sub-environments,
     b) or the survival of population (which "wants" to  transit)
is more then the survival some population in new environment.
     9.1.  We  can  simulate  also  additional   difficulty   for
transitions. When survival of population which "want" to  transit
decreases on some quantity (special parameter).
     10. Competition into environment: in each cycle of  work  of
program the population which have minimal survival is replaced by
population  which  have  maximal  survival  (this  is  into  each
environment). The  replacement  must  occur  if  there  are  N>=2
populations  in  an  environment,  and  if   survivals   of   the
populations are not equal.
     10.1.  "Non  hard  competition".  This  is  coexistence   of
populations which have survival differed on  the  some  quantity.
The simulation of competition is not produced if survivals differ
from each to other less then some quantity (special parameter).
     11. If any region in any  environment  in  empty,  then  the
population which has the largest "summa" of  all  populations  in
all environments will colonise the empty region.
     12. At the beginning we put usually one population into some
one environment.
     13. The program work by cycles. The description  of  job  of
all acts in cycle is below.
     The first cycle: we put one population in some  environment.
After this in each cycle we have:
     a) mutations,
     b) abolishing  of  populations  which  have  non  compatible
features,
     c) calculation of survivals for all  populations  into  each
environment,
     d) transitions (if it is possible),
     e) competition in each environment,
     f)  filling  of  all  free  positions  into  partial  filled
environment  (note:  only  into  environments  into  which   some
quantity of already populations exist),
     g) return to a),
     etc.

     The simulation, hence,  models  certain  basic  evolutionary
phenomena:
     1. Mutation.
     2. Negative selection of non-viable individuals.
     3. Migration and colonization processes.
     4. Competitive exclusion of a species by another species.
     Computer   experiments   show   that   some    peculiarities
of competitive interrelation are very important. In particularly,
the capacity of  coexistence  of  populations  having  difference
survival in one environment is essential. The special  parameters
have included in the model to investigate similar effects.

                    THE RESULTS AND DISCUSSION

     Since  we  were   looking   at   large-scale   patterns   in
phylogenetic evolution,  we  needed  to  chose  an  appropriately
high-level of taxonomic grouping for our study. We  selected  the
phylum Chordata to provide data and  to  give  us  a  comparative
phylogeny with which we might evaluate our  results.  Of  course,
this programing complex suits completely for many other (and  non
biological even) objects.
     The "Mutations," hence, in our modeling correspond to  major
changes inadaptive traits. They represent adaptive  breakthroughs
such as major changes in size, or in the nature  of  respiration.
In nature, of course, the actual time taken for such events is on
the order of .5-1 million years.
     It should be noted that these are "Macromutational"  changes
and the model therefore deals with the consequences of such major
adaptive shifts, rather than their causes.
     Our modeling of the environment was handled in  a  similarly
broad manner: for example we posited the existence of 4  aquatic,
3 fresh water, and 3 land  environments.  To  this,  we  added  2
environments in which an amphibial adaptation would be optimal.
     We began the simulation of evolutionary change in our  model
with a population having characteristics similar to  Lanceolatus.
At each iteration of the program, processes of macromutation  and
migration were simulated. Hence, given the rate for the evolution
of these  kinds  of  adaptive  change,  each  iteration  was  the
equivalent of between .5 and  1  million  years.  Generally,  the
simulation was carried  out  through  40  -  60  cycles,  thereby
modeling evolutionary  events  occurring  over  periods  of  some
200-300 million years.
     The simulation of macro-evolutionary patterns  studied  here
demonstrated   that   certain   unexpected   effects,   such   as
"Reimigration of near descendants" and some  others  (see  below)
may  be  observed  in  the  patterns  produced  by   evolutionary
processes. This result demonstrates  that  the  "Reimigration  of
near descendants" is the significant factor of evolution.  If  to
say by another words, the results of the simulation  demonstrated
that patterns of allopatric speciation  followed  by  competitive
replacement in the original  region  is  a  common  pattern  over
evolutionary time and has major effects upon the  nature  of  the
evolutionary trees which the program generates.
     Other interesting results have been obtained in  experiments
concerning to the change of parameters of competition. We uncover
hard  competition  provokes  such  processes  when  the  ones  of
specialization predominate  processes  of  filling  out  of  free
environment. This  process  appears  to  be  a  function  of  the
mutation  rate.  More  exactly:  the  increase  of  frequency  of
mutations promotes in the intensification of this tendency.
     In result of this experiments we obtain that in case of hard
competition the lateral branches of evolutionary trees don't grow
almost and are the  trunk  composed  from  biological  forms  the
survival of them increases gradually. Many environments  remained
unfilled. In biological terminology  this  is  means  that  under
these conditions, the amount of diversification is  minimized  as
is the rater at  which  new  habitats  are  colonized.  Decreased
competition  into  environments  also  tends  to   increase   the
diversification and the rate of migration and colonization!
     The simulation also allows us the opportunity to investigate
how evolutionary trees constructed from a sub-set  of  the  total
number of traits differ from those in which all trait states  are
know. Construction of evolutionary trees from  partial  data,  of
course, is the rule in biology,  with  areas  of  study  such  as
paleontology, genetics and evolutionary physiology having to work
with less than complete knowledge of the total  trait  assemblage
of the organisms studied.
     Since the simulation contains a "total"  data  set,  we  may
then  easily  compare  the   "true"   phylogenies   with   others
constructed from incomplete data. Here,  we  followed  the  usual
convention of establishing  ancestor  -  descendant  and  "sister
species" relationships by minimizing the number of modified trait
states.  We  found  that  trees  reconstructed  from   incomplete
information  differed  significantly  from  the   "true"   trees.
Moreover,  this  is  means  that  trees  are  built  for  various
biological  sciences  need  not   consider   as   ones   differed
paradoxically but can  be  described  as  trees  completing  each
other.
     All  in  all,  however,  the  model  of   macro-evolutionary
processes produced by our program, tends to confirm certain basic
assumptions generally used in reconstructing the  paleontological
record.
     1. The probability of return to ancestral forms is very low;
evolution is non reversible.
     2. The overall pattern of branching in evolution is a result
of random changes, which have not significance  for  survival  of
species population.
     3.   Some   peculiarities   of   ecological   evolution   is
predetermined. This is the consequence of  the  same  results  in
particularity  concerning  to  filling  of  environments  may  be
obtained in case of various parameters  of  model  (frequency  of
mutations for instance).
     4. Highly  specialized  species  seldom  produce  descendant
species.

                          REFERENCE LIST

Levchenko  V.F.  "A  physic-ecological  conception  of  biosphere
evolution."  [in  Russian]  In  Proceeding  of   the   1990   7th
International Symposium "Simulation  of  system  in  biology  and
medicine" (Praha), 59-63.

Levchenko V.F. 1992."Directedness of biological  evolution  as  a
consequence of the biosphere  development  [in  Russian]  Zhurnal
obshchei biologii Vol. 53, no 1(Jan-Feb): 58-70.

Levchenko  V.F.  and  Menshutkin  V.V.   1988.   "Simulation   of
macroevolutionary  process"  Zhurnal  evolucionnoi  biokhimii   i
fiziologii. Vol. 23, no. 5(Sep-Oct): 668-673.

Levchenko V.F. and Starobogatov Ya.I.  1990  "Succession  changes
and evolution  of  ecosystems."  [in  Russian]  Zhurnal  obshchei
biologii Vol. 51, no 5(May): 619-631.

