1: \documentstyle[aps,multicol,epsfig,eqsecnum]{revtex}
2:
3: \begin{document}
4: \draft
5:
6: \title{Evolution of polymorphism and sympatric speciation through competition
7: in a unimodal distribution of resources}
8:
9: \author{E. Brigatti $^{\dag\pm}$,
10: J.S. S\'a Martins $^{\star}$
11: and I. Roditi $^{\dag}$
12: }
13:
14: \address{$\dag$Centro Brasileiro de Pesquisas F\'{\i}sicas, Rua Dr. Xavier
15: Sigaud 150, 22290-180, Rio de Janeiro, RJ, Brasil}
16: \address{$\star$Instituto de F\'{\i}sica, Universidade Federal Fluminense,
17: Campus da Praia Vermelha, 24210-340, Niter\'oi, RJ, Brasil}
18: \address{$\pm$e-mail address: edgardo@cbpf.br}
19:
20: \maketitle
21:
22: \widetext
23:
24: \begin{abstract}
25: A microscopic agent dynamical model for diploid age-structured populations
26: is used to study evolution of polymorphism and sympatric speciation. The
27: underlying ecology is represented by a unimodal distribution of resources of
28: some width. Competition among individuals is also described by a similar
29: distribution, and its strength is maximum for individuals with the same
30: phenotype and decreases
31: with distance in phenotype space as a gaussian, with some width. These two
32: widths define the model's phase space, in which we identify the regions where
33: an autonomous emergence of stable polymorphism or speciation is more likely.
34:
35: \end{abstract}
36:
37: \pacs{87.23.Kg, 87.23.-n, 05.10.Ln}
38:
39: %\begin{document}
40:
41: %\maketitle
42:
43: \begin{multicols}{2}
44:
45: \section{Introduction}
46:
47: The genesis of the incredible diversity of life on our planet is the problem
48: that lies at the heart of the development of modern theory of evolution. This
49: theory attempts to describe all its diverse processes, and in particular
50: speciation, as the outcome of a microscopic dynamics driven by selection and
51: mutation. Insofar as the biological definition of species goes, speciation is
52: the process that explains the generation of two reproductively isolated
53: populations, for which gene flow between the different taxa is absent in any
54: form. The most intuitive conjecture as to how speciation can occur leads to a
55: scenario in which a geographical barrier, or any physical isolation
56: mechanism, separates some fraction of the population of a species. This
57: process is called allopatric speciation. This description is nowadays well
58: accepted and supported by an abundance of empirical evidence \cite{alop,Coyne}.
59: Much more subtle and complex is the conjecture that speciation is also
60: possible in the absence of any physical isolation mechanism: a single
61: continuous population of interbreeding organisms can split into two
62: reproductively isolated subpopulations. This conjecture is not new, dating at
63: least from Darwin himself \cite{Charles}, albeit from a perspective that is
64: somewhat different from the modern one. A great interest in this theme has
65: grown lately \cite{Maynard,Lande}, strongly supported by observational data related
66: to some fishes of small crater lakes of Cameroon \cite{Schliewen} and by
67: recent papers that demonstrate sympatric speciation in vitro \cite{Friesen}
68: or in artificial-life type simulations \cite{Chow}. On the other hand, the
69: effort towards the construction of general models that can give structure and
70: plausibility to verbal theories (see Ref. \cite{Coyne,Turelli} for a general
71: review) have caused the investigation of an extensive number of mathematical
72: and computational models. Out of the wide variety of such models, we restrict
73: our analysis to ecology-driven reproductive isolation, and we address our
74: attention to models in which natural selection depending on resources
75: distribution originates sympatric speciation. Some of the early attempts on
76: these lines were made with models that represent heterogeneous environments
77: with two niches, and where a mechanism of adaptation to such discrete
78: resources was established. In this case, speciation is caused by a fixed
79: selection that favors phenotypes of both the extremes of the possible range
80: (disruptive selection) and leads to reproductive isolations. A model with
81: these ingredients appeared recently in the literature \cite{KK}, and a
82: similar one, based on microscopic dynamics, was studied by one of the authors
83: \cite{Jorge,Karen}. The reasoning that underlies these models suggests the
84: question of whether discrete niche might be necessary to explain species
85: coexistence in sympatry. A recent analysis \cite{Dieckmann} tried to give a
86: negative answer to this question, showing how speciation could arise from
87: competition for continuously distributed resources. In such situations, the
88: environment has just a single ecological niche and, as a consequence, the
89: population feels a stabilizing selection that favors intermediate phenotype.
90: In this scenario, the introduction of a competition mechanism generates a
91: disruptive selection that weakens as species diverge. This competition is
92: generated through a selection force that is stronger for more frequent
93: phenotypes and weaker for rare ones (frequency-dependent selection).
94: This mechanism is interesting for two different reasons. First, such a sequence
95: of events is controlled by a
96: frequency-dependent interaction and does not require any externally imposed
97: disruptive selection pressure. On top of that, it appears to be a more
98: general mechanism, which can occur under a much wider range of ecological
99: situations. These two selection pressures are yet not enough to generate
100: speciation. To obtain a stable process, it is necessary to prevent the
101: appearance of intermediate phenotypes through the divergence of reproductive
102: compatibility. This is obtained thanks to the evolution of assortativity,
103: a non-random mating strategy.
104: In the case of a unimodal distribution of resources, such a mechanism is also
105: strictly necessary for the appearance of a first phenotypic
106: divergence. All these ideas have been recently developed in Ref.
107: \cite{Dieckmann} and are the point of departure for our present analysis,
108: where we want to study all these results for a more complex microscopic model.
109:
110: \section{Model and methods}
111:
112: We are interested in working at a microscopic level, where the fundamental
113: agent of the model is an individual of an age-structured population with
114: phenotype variability. For this reason, the classical Penna model
115: \cite{Penna} is a convenient point of departure, in its asexual haploid and
116: sexual diploid versions. In the following, we will describe results relative
117: to both sexual and asexual taxa. In the first case, we are dealing with the
118: onset of sympatric speciation in a strict sense, which coincides with the
119: setting up of reproductive isolation. In the second, we are investigating the
120: emergence of polymorphism, meaning the splitting of a phenotypical
121: monomorphic population into two (or more) distinct phenotypic clusters. The
122: sexual case describes a diploid population, representing each individual
123: through two sets of coupled genetic strands. Each one is built up by two
124: $32$-bit long bit-strings. The first string is age-structured and is used to
125: introduce the biological clock of the individuals, while the other represents
126: the phenotype (trait bit-string). In each of these strings, a gene is encoded
127: by two homologous bits. Its value is $0$ if it describes an allele equal to
128: the one present in the original population - wild type allele - or $1$ if it
129: has suffered a change and has become a mutant allele. Each individual of a
130: reproductively active couple generates, through a meiotic cycle with
131: crossing-over and recombination, one haploid gamete that, after the
132: introduction of some mutations, combine to form a new genetic strand (see Fig.
133: \ref{fig_Exa}).
134:
135: The phenotype value is associated with the second string of the genome by the
136: overall sum of the active mutations present in that portion. By active
137: mutation we mean a homozygous locus, where the bit value of the two
138: homologous alleles are equal to one, or a heterozygous one for which the
139: mutant allele is dominant. The number of loci where the mutant alleles are
140: dominant is fixed and their position is chosen randomly at the beginning of the
141: simulation. According to this procedure, the phenotype value
142: ($x$) is an integer value between $0$ and $32$, and is different from the
143: simple sum of all the ones present in the bit-string. This kind of
144: characterization of the phenotype takes into account the diploid nature of
145: the genome and is in some way consistent with the representation of a
146: quantitative trait. This must be understood in the sense that we are
147: representing a trait depending on multiple genes (polygenic trait), but in a
148: particularly simplified situation: the environment has a negligible effect on
149: the phenotype and the quantitative trait is ideal (the genes are randomly
150: associated and their effect is completely additive \cite{Daniel}).
151:
152: The genetic and, in correspondence, phenotypic variability is assured by the
153: action of mutations. Mutations are obtained by randomly flipping one of the
154: bits of the string. Bits of the trait bit-strings can mutate from $0$ to $1$
155: or from $1$ to $0$, as opposed to the ageing bit-string that can undergo only
156: bad mutations. This is an oversimplification: for the trait bit-string we do
157: not take into account that it is really difficult for a mutated allele to
158: restore its previous activity. On the contrary, for the ageing bit-string we
159: consider only harmful mutations because of their strong predominance in
160: nature. With the use of this phenotype characterization it is possible to
161: establish a computational representation of intraspecific and/or
162: environmental interaction and/or sexual selection \cite{Jorge,Karen,Edgardo}.
163:
164: To obtain an age structure in the population, we allow each agent to live
165: until the occurrence of death caused by ageing. This is implemented following
166: rules inspired by Medawar's hypothesis of the accumulation of bad mutations. A
167: position (locus) of the chronological (age-structured) piece of the genome is
168: read at each time step. If an active mutation (defined in the same manner as
169: for the phenotype value) is found at this locus, it is added to the current
170: number of harmful mutations; the individual dies when this amount reaches
171: some pre-determined threshold value.
172:
173: In the model we introduce another death factor that represents the natural
174: causes of selection in a real ecosystem. It is characterized by three
175: different components. The first is density-dependent, responsible for
176: limiting the number of the total population in accordance with a logistic
177: growth. Then, a frequency-dependent factor takes into account how, in
178: realistic situations, the tendency to occupy the more favored regions in
179: phenotype space is contrasted to an increasing competition between
180: individuals. To these dynamic components of selection, which represent the
181: feedback between individuals and ecosystem and take into account the
182: evolution of the environment, a static component is added. It designs the
183: general ambient condition of the territory and the ecological niche where the
184: agents live. This is a cause of a directional selection that constantly
185: drives the population towards a fitness maximum in phenotype space. To sum
186: up, the general expression for such death factor, as used
187: in the Monte Carlo simulations, is:
188: \begin{eqnarray}
189: V = \frac{\sum_{y=1}^{32} N_{y}\cdot\exp(-(x-y)^2/2c^2)}
190: {k \cdot \exp(-(x-16)^2/2s^2)}\nonumber
191: \end{eqnarray}
192: At each time step in the simulation a random number is tossed; the individual
193: survives if this number is larger than the $V$ value. The static fitness
194: landscape is represented by a gaussian with deviation $s$ and drives all the
195: population towards the fitness maximum equal to $16$. The competition
196: declines with phenotype distance according to a gaussian function with
197: deviation $c$, and the parameter $k$ controls the population dimension. $x$ is
198: the phenotype value of the individual that is feeling the selection pressure
199: and the sum runs over the $y$ index that spans all of the phenotype space.
200: With $N_{y}$ we indicate the number of individuals with phenotype $y$.
201:
202: As already mentioned, positive assortative mating plays a basic role in the path
203: towards speciation. This is
204: obtained by allowing mating only if the distance in phenotype space between
205: two partners is smaller than some predefined threshold $a$ \cite{Franco}.
206: It is interesting to notice that several models (for example in \cite{Dieckmann})
207: implement some mechanism that, by normalizing mating probabilities, guarantees
208: an equal reproductive success for all individuals. This care is due to the fact
209: that assortative mating can cause a sexual selection on rare phenotypes, that have
210: a smaller possibility of finding mates. This byproduct of assortative mating,
211: which has a stabilizing effect that makes speciation less likely
212: \cite{Kirkpatrick}, can be present in nature. For this reason, in our model, we do
213: not implement any mechanism for equalizing mating success.
214:
215: Finally, we are not focusing on the problem of the natural evolution of
216: assortativity
217: \cite{Dieckmann}; in contrast, we are assuming here a fixed assortative rule
218: from the start, for the sake of simplicity. In fact, with such a schematic
219: mechanism, the tuning of the parameter that leads to reproductive isolation
220: is a hard problem.
221:
222: \section{Results}
223:
224: We begin the presentation of our results by showing some general data for an
225: asexual population. The dynamical rules for such a population are exactly the
226: same as for the sexual version, except for the fact that we are now dealing
227: with a haploid population, with a very simple reproductive cycle, and where
228: the mutation process stands as the only cause of variability \cite{Penna}.
229:
230: In this simple case the population evolves rapidly towards the fittest
231: phenotype ($16$) that corresponds to the maximal carrying capacity. Once this
232: value has been reached, two scenarios are possible. In the first, $16$ is a
233: stable point in phenotype space and the population will structure itself in a
234: well shaped gaussian distribution around such value. Otherwise, $16$ is still
235: not a fitness maximum (see figure \ref{fig_Dyn}), and the population
236: experiences disruptive selection that leads to the appearance of two distinct
237: phenotypic morphs. In this latter case, the tendency to avoid an overcrowded
238: region of phenotype space is stronger than the advantage obtained by reaching
239: the optimal phenotype value. This happens when $c<s$ (see table
240: \ref{tab_Asex}). This is not the only possible behavior though; for some
241: small enough values of $c$, the population rapidly splits up its phenotype
242: distribution before its mean reaches the fittest phenotype. An example of
243: this situation is given in the sexual case (see Fig. \ref{fig_Dyn3}). This
244: particular behavior is related to a stronger influence of competition, that
245: drives a premature bifurcation. In fact, this force increases as $c$ is
246: decreased, as also reflected in Table \ref{tab_Asex} by a larger number of
247: occurrences of branching events. This is easily seen, for instance, when
248: $c = 10$ where, in just $100000$ time steps the population undergoes two
249: bifurcation events, each one causing the appearance of two distinct phenotype
250: clusters, leading towards a three-modal distribution. As we increase the
251: competition further, the population feels such a strong drive that it is
252: impossible to reach a stable polymorphism: the branching events are now so
253: numerous that the phenotype distribution becomes unstable, characterized by a
254: large number of peaks connected by intermediate phenotypes (see inset in
255: figure \ref{fig_Dyn}). These results are comparable with the ones obtained in
256: Ref. \cite{Dieckmann}, but the assignment of the phenotype from a genome ruled by
257: microscopic dynamics determines a richer phenomenology.
258: In fact, their prediction of branching for
259: $c<s$ is confirmed by our study, but we were also able to show how branching
260: can occur repeatedly, leading to a polymorphism with more than two phenotype
261: clusters, or to even more complex unstable behavior, as reflected in the
262: enriched unfolding of our model in time.
263: Moreover, data obtained by longer simulation show how the clustering of the
264: population phenotype is biological relevant. In fact, also after $10^6$ time steps
265: polymorphism is still present (see figure \ref{fig_Long}). For this reason, even if from a numerical
266: simulation it is impossible to determine the final equilibrium distribution, such a
267: long-lived structure can not be considered a simple transient phase
268: \cite{Polechova} but, at least, as some sort of long-standing quasi-stationary
269: state with a time scale comparable to the ones of biological interest.
270:
271: Evolution of diploid sexual populations is marked by a stronger variability,
272: due to the effects of Mendelian segregation and recombination. This fact is a
273: source of difficulties in obtaining sharp phenotypic differentiations because
274: of the constant generation of intermediate phenotypes. For this reason, if the
275: mating process is random, a stable splitting into two different phenotypic
276: clusters can no longer occur, for any possible value of $s$ and $c$. On the
277: contrary, with the introduction of positive assortative mating, not only it is
278: possible to recognize different phenotypic morphs but also speciation with
279: sexually isolated populations (Fig. \ref{fig_branch}). The parameter space
280: where it occurs appears to be more confined than in the asexual cases.
281: Although still necessary, it is no longer sufficient that $c<s$. Out of the
282: various simulations performed, we have been able to obtain speciation for
283: some simulations with parameters set in the following region:
284:
285: \begin{eqnarray}
286: s &\ge& 15, 6 \le c \le 8, a=4;\nonumber \\
287: s&=&40, c=6, 3 \le a \le 5 \nonumber
288: \end{eqnarray}
289:
290: \noindent
291: The simulations show that the system is not very sensitive with respect to
292: the value of $s$. If it is sufficiently larger than a particular value of
293: $c$, the population is able to escape from the region with maximum carrying
294: capacity. On the other hand, the value of $c$ can neither be too small, which
295: would lead to a distribution characterized by not well defined connected
296: peaks, nor too large, otherwise the competition would not be strong enough to
297: drive the distribution away from a gaussian centered in $16$. Finally, values
298: of $a$ that are too small can totally trap the variability of the phenotype
299: distribution to a confined region close to $0$ or even prevent speciation because
300: of the stronger penalizing effect of sexual selection on rare phenotypes
301: \cite{Kirkpatrick}.
302: For larger values, on the
303: opposite, the mating choice is not strict enough to prevent intermediate
304: phenotypes. In any case, even in this more restricted scenario, it is still
305: possible to find some more favorable situations for which, after just
306: $100,000$ time steps, two speciation events have already occurred (see figure
307: \ref{fig_Dyn3}). With respect to the dependence on the initial condition, we
308: remark that if we start with a population having a phenotype close to $16$ it
309: is, in general, easier to obtain speciation. We again claim our results to be
310: in accordance with those of Dieckmann and Doebeli \cite{Dieckmann}, but the
311: parameter requirement for the emergency of speciation appears to be more
312: restrictive, even with the usage of a stronger and fixed mating rule. We have
313: also performed some simulations with a model characterized by a big trait
314: bit-string (up to $160$ loci) \cite{Polechova}. The fact that no qualitative
315: differences emerged suggests that the results obtained are not dependent on the
316: phenotype range (see Fig. \ref{fig_branch160}).
317:
318: \section{Conclusions}
319:
320: The representation of an individual phenotype character through a
321: second bit-string is a natural extension of the classical Penna model that allows
322: the study of sympatric speciation. Our model was structured under these
323: guidelines and the results corroborate the hypothesis that sympatric speciation
324: emerges from competition for continuously distributed resources. We also used it
325: to test the robustness of this hypothesis under the inclusion of further
326: ingredients and the dependence of the results on the values of some of the model's
327: constitutive variables.
328:
329: With the study of an age-structured population we can substantiate the claim that
330: the precedent results were not biased by possible long-lived individuals.
331: Moreover, in the asexual population the introduction of a phenotype determined by
332: a genotype with $32$ loci confirm previous results \cite{Dieckmann} in an enriched
333: scenario. The fact that clustering is still present, even after more than $10^6$
334: time steps, determines the biological relevance of such a long-lived phase.
335: For a sexual population where the mechanisms of crossing-over, dominance and
336: homo/heterozygous loci were implemented, the emergence of sympatric speciation is
337: obtained. Our results, obtained with a model that does not include any mechanism
338: for equalizing mating success, contradict the alleged stabilizing effect due to
339: sexual selection caused by assortative mating. Finally, the agility of the Penna
340: model in manipulating bit-strings allowed the run of simulations with large
341: phenotype range, with results that suggest that there is no dependence between
342: speciation and trait range, in spite of a recent claim to the contrary
343: \cite{Polechova}. It is our opinion that agent-based models that allow statistical
344: fluctuations, such as the one in Ref. \cite{Dieckmann} and ours, are the most
345: promising testing grounds for evolutionary ideas as a whole.
346:
347: \section*{Acknowledgments}
348:
349: We thank the Brazilian agencies CAPES, CNPq, and grants from PRONEX
350: (PRONEX-CNPq-FAPERJ/171.168-2003 and PRONEX-FAPERJ E-26/171/2003) and FAPERJ
351: (E-26/170.699/2004) for partial financial support.
352:
353: \begin{thebibliography}{99}
354:
355: \bibitem{alop}
356: W.R. Rice,and E.E. Hostert, Evolution {\bf 47}, 1637 (1993).
357:
358: \bibitem{Charles}
359: C. Darwin, {\it On the origin of species}, Cap.4, J. Murray, London (1859).
360:
361: \bibitem{Coyne}
362: J.A. Coyne and H. A. Orr, {\it Speciation}, Sinauer Associates,
363: Sunderland, MA (2004).
364:
365: \bibitem{Maynard}
366: J. Maynard Smith, Am. Nat. {\bf 100}, 637 (1966).
367:
368: \bibitem{Lande}
369: R. Lande, Proc. Natl. Acad. Sci. USA {\bf 78}, 3721 (1981).
370:
371: \bibitem{Schliewen}
372: U.K. Schliewen, D.Tautz and S. P\"a\"abo, Nature {\bf 368}, 629 (1994).
373:
374: \bibitem{Friesen}
375: M.L. Friesen et al., Evolution {\bf 58}, 254 (2004).
376:
377: \bibitem{Chow}
378: S.S. Chow, C.O. Wilke, C. Ofria, R.E. Lenski and C. Adami, Science {\bf 305},
379: 84 (2004).
380:
381: \bibitem{Turelli}
382: M. Turelli, N.H. Barton and J.A. Coyne, Trends in Ecol. \& Evol. {\bf 16},
383: 330 (2001).
384:
385: \bibitem{KK}
386: A.S. Kondrashov and F.A. Kondrashov, Nature {\bf 400}, 351 (1999).
387:
388: \bibitem{Dieckmann}
389: U. Dieckmann and M. Doebeli, Nature {\bf 400}, 354 (1999).
390:
391: \bibitem{Penna}
392: T.J.P. Penna, J. Stat. Phys. {\bf 78}, 1629 (1995).
393:
394: \bibitem{Kirkpatrick}
395: M. Kirkpatrick and S.L. Nuismer, Proc. R. Soc. Lond. B {\bf271}, 687 (2004).
396:
397: \bibitem{Daniel}
398: D.L. Hartl, {\it A primer of population genetics}, Sinauer Associates,
399: Sunderland, MA (1988).
400:
401: \bibitem{Jorge}
402: J.S. S\'a Martins, S. Moss de Oliveira and G.A. de Medeiros,
403: Phys. Rev. E {\bf 64}, 021906 (2001).
404:
405: \bibitem{Karen}
406: K. Luz-Burgoa, S. Moss de Oliveira. J.S. S\'a Martins, D. Stauffer
407: and A.O. Sousa, Braz. Jour. Phys. {\bf 33}, 623 (2003).
408:
409: \bibitem{Edgardo}
410: E. Brigatti, J.S. S\'a Martins, I. Roditi, Eur. Phys. J. B {\bf 45}, 529 (2005).
411:
412: \bibitem{Franco}
413: F. Bagnoli and C. Guardiani, Physica A {\bf 347}, 534 (2005).
414:
415: \bibitem{Polechova}
416: J. Polechova and N.H. Barton, Evolution, {\bf 59}, 1194 (2005).
417: \end{thebibliography}
418:
419: \newpage
420:
421: \begin{table}[p]
422: \begin{center}
423: \begin{tabular}{cc|ccccccc}
424: %\hline
425: & 40 & 1 & 1 & 1 & 1 & 1 & 1 & 1\\
426: {\bf $c$} & 35 & 1 & 1 & 1 & 1 & 1 & 1 & 2\\
427: & 30 & 1 & 1 & 1 & 1 & 1 & 2 & 2\\
428: & 25 & 1 & 1 & 1 & 1 & 2 & 2 & 2\\
429: & 20 & 1 & 1 & 1 & 2 & 2 & 2 & 2\\
430: & 15 & 1 & 1 & 2 & 3 & 3 & 3 & 2\\
431: & 10 & 1 & 3 & 3 & 3 & 3 & 3 & 3\\
432: & 5 & x & x & x & x & x & x & x\\
433: %\hline
434: \hline
435: & & 10 & 15 & 20 & 25 & 30 & 35 & 40\\
436: & & & & & & & {\bf $s$} & \\
437: \end{tabular}
438: \end{center}
439: %\noindent
440: \caption{ \small Generation of polymorphism in an asexual population as we vary
441: the values of the standard deviation of the competition ($c$) and of the static
442: component ($s$). Each number in the table corresponds to the number of isolated
443: peaks in the phenotype distribution, while a cross indicates the occurrence of
444: unstable polymorphism, such as the one represented in the inset in figure
445: \ref{fig_Dyn}. These results are the outcome of one simulation.
446: We have performed a number of different simulations varying the initial condition.
447: In all cases, the appearance of distinct phenotypic morphs happens only when
448: $c<s$, and the overall dependence of the results on the parameters is similar,
449: even when the number of peaks, isolated or not, vary quite a lot.
450: The simulations run with the same parameters as those of figure \ref{fig_Dyn}.
451: The picture was taken after $100,000$ time steps, after all
452: distributions have reached a stationary situation.}
453: \label{tab_Asex}
454: \end{table}
455:
456: \begin{figure}[p]
457: \begin{center}
458: \vspace*{0.8cm}
459: \resizebox{0.4\textwidth}{!}{\includegraphics{phen2.eps}}
460: %\centerline{\psfig{figure=phen2.eps, width=8.5cm, angle=0}}
461: \vspace*{0.4cm}
462: \end{center}
463: \caption{\small Example of a reproductive cycle.
464: The diploid genome is represented
465: with its age-structured part (light-shaded background)
466: and the bit-strings that encode
467: for phenotype (diagonal stripes background).
468: After performing crossing-over in the first passage,
469: the haploid gamete is chosen in the second.
470: Finally, some new mutation are added
471: (dark-shaded squares) and the gametes
472: combine to form a new individual.}
473: \label{fig_Exa}
474: \end{figure}
475:
476: \begin{figure}[p]
477: \begin{center}
478: \vspace*{0.8cm}
479: \resizebox{0.4\textwidth}{!}{\includegraphics{dynamic25-20bis.eps}}
480: %\centerline{\psfig{figure=dynamic25-20bis.eps, width=8.5cm, angle=0}}
481: \vspace*{0.4cm}
482: \end{center}
483: \caption{\small Time evolution of an asexual population. The parameters used
484: in the simulations are: $k$ ($100,000$), the initial population ($1,000$),
485: the minimum reproduction age ($6$), the maximum reproduction age ($32$), the
486: number of offspring per mating season ($2$), the threshold value for harmful
487: diseases ($3$), the number of mutations added at birth in the age structured
488: string ($1$), and the mutation probability in the trait string ($0.01$). The
489: values for the standard deviation of the competition ($c$) and of the static
490: component ($s$) are respectively $20$ and $25$. The inset shows the unstable
491: distribution generated by a simulation with $s=25$ and $c=5$ after $100,000$
492: time steps.}
493: \label{fig_Dyn}
494: \end{figure}
495:
496: \begin{figure}[p]
497: \begin{center}
498: \vspace*{0.8cm}
499: \resizebox{0.4\textwidth}{!}{\includegraphics{Longtime.eps}}
500: %\centerline{\psfig{figure=Longtime.eps, width=8.5cm, angle=0}}
501: \vspace*{0.4cm}
502: \end{center}
503: \caption{\small The standard deviation of the phenotypes distribution can be
504: considered as an order parameter: the splitting of a phenotypical monomorphic
505: population into two (or more) distinct phenotypic clusters corresponds to an
506: abrupt increase in the value of the standard deviation. The fact that after
507: the first transition, near $t=30000$, the mean value of the standard
508: deviation does not change anymore proves the stability of the bimodal
509: phenotypic distribution ($s=25$, $c=20$).}
510: \label{fig_Long}
511: \end{figure}
512:
513: \begin{figure}[p]
514: \begin{center}
515: \vspace*{0.8cm}
516: \resizebox{0.4\textwidth}{!}{\includegraphics{branch.eps}}
517: %\centerline{\psfig{figure=branch.eps, width=8.5cm, angle=0}}
518: %\resizebox{0.4\textwidth}{!}{\includegraphics{branch.eps}}
519: \vspace*{0.4cm}
520: \end{center}
521: \caption{ \small An example of speciation in a sexual population. The figure
522: shows the phenotype distribution after $100,000$ time steps. The parameters
523: used in the simulations that are different from those of Figure \ref{fig_Dyn} are:
524: $s=40$, $c=8$, $a=4$, the minimum reproduction age ($8$), the number of
525: offspring per mating season ($4$), and the number of loci where the $1$ allele
526: is dominant in each string ($16$).}
527: \label{fig_branch}
528: \end{figure}
529:
530: \begin{figure}[p]
531: \begin{center}
532: \vspace*{0.8cm}
533: \resizebox{0.4\textwidth}{!}{\includegraphics{branch160.eps}}
534: %\centerline{\psfig{figure=branch160.eps, width=8.5cm, angle=0}}
535: \vspace*{0.4cm}
536: \end{center}
537: \caption{ \small A simulation with a trait
538: bit-string of $160$ loci ($s=160$, $c=8$, $a=4$).
539: This result suggests that there is no dependence between speciation and trait
540: range, in spite of a recent claim to the contrary [19]. }
541: \label{fig_branch160}
542: \end{figure}
543:
544: \begin{figure}[p]
545: \begin{center}
546: \vspace*{0.8cm}
547: %\centerline{\psfig{figure=dynamic3.eps, width=8.5cm, angle=0}}
548: \resizebox{0.4\textwidth}{!}{\includegraphics{dynamic3.eps}}
549: \vspace*{0.4cm}
550: \end{center}
551: \caption{\small A particular situation where it is possible to rapidly obtain
552: two speciation events ($s=40, c=6, a=4$). }
553: \label{fig_Dyn3}
554: \end{figure}
555:
556: \end{multicols}
557:
558: \end{document}
559:
560: