Introduction

The very early forms of life have been suggested to comprise RNA-like molecules that served as templates for their own replication and catalyzed various reactions (see e.g. Dworkin et al. 2003; Orgel 2004). Experimental evidence supports these hypotheses as it has been demonstrated that two RNA replicating enzymes (i.e. ribozymes) can catalyze each other’s synthesis by using RNA as a template (Lincoln and Joyce 2009). The ancient analogs of these types of molecules probably originated in an environment where components for their synthesis concentrated spontaneously and were thus readily available. Mid-ocean hydrothermal vents have been hypothesized as such early habitats (Baross and Hoffman 1985; Maher and Stevenson 1988). Within the vents there are tiny compartments that could have served as abiotic cells for primordial life (see e.g. Koonin and Martin 2005) and simulations have shown that these compartments may accumulate nucleotides (Baaske et al. 2007). It was demonstrated recently that the pyrimidine ribonucleotides can be synthesized in prebiotically plausible conditions (without the free ribose, free nucleobase stage) through arabinose amino-oxazoline and anhydronucleoside intermediates, which thus promotes the credibility of the presented hypotheses about the early forms of life (Powner et al. 2009). Furthermore, it is possible that the early evolution of the RNA-world took place within membrane-surrounded vesicles (Hargreaves et al. 1977; Monnard and Deamer 2001; Monnard and Deamer 2002; Furuuchi et al. 2005).

There are many hypotheses that involve a dynamic, primordial gene-community. For example, the life forms that existed before the last universal common ancestor (LUCA) were probably members of a community where horizontal gene transfer was frequent (Woese 1998; Woese 2002). The early emergence of viruses could also be situated into such a community (Iyer et al. 2006; Koonin et al. 2006) and the universal genetic code might be a direct result of the early, communal selection for innovation-sharing protocols (Vetsigian et al. 2006). Virus-like parasites could have helped to share information between the abiotic cells and later on the selfish parasites might have promoted the isolation of Bacteria, Archaea, and Eukarya into individual, vertically evolving domains (Forterre 2006; Jalasvuori and Bamford 2008).

Computational studies concerning the origin and the evolution of early replicators, multiple templates and, in some cases, dividing vesicles have been conducted (Eigen and Schuster 1979; Hogeweg and Takeuchi 2003; Stadler and Stadler 2004; Szathmáry 2006; Fontanari et al. 2006). Theoretical studies have treated questions such as the requirements of replication fidelity of replicators (Eigen 1971), the rate of decay of replicators (Szathmáry 2006), and the increase of complexity in RNA-replicator systems (Takeuchi and Hogeweg 2008). In some studies the dynamics of replicators are observed in proto-cells that are able to divide either randomly, depending on the genetic content or due to mechanical stress (Szathmáry and Demeter 1987; Zintzaras et al. 2002; Fontanari et al. 2006). However, few studies consider the evolutionary dynamics of a community of replicators and other simple genetically encoded agents within a static compartment matrix. Therefore, in this study, a virtual computer-based system designed to study the evolution of a community of ribozyme-like agents is presented. The system is based on rules in a similar manner to some of the earlier computing systems that have treated the evolution of bacterial populations (Vlachos et al. 2006; Gregory et al. 2008).

In this study the general events that take place in a typical simulation were observed and then experiments were performed to answer some (previously presented) hypotheses. The results indicated, among other things, that active sharing of information between compartments is important for evolution towards more complex systems and that the presence of parasites cause the selection of anti-parasite genes.

Virtual Computing System

A virtual computing system, entitled as PrimordialEvo, was developed to simulate a community of ribozyme-like agents. The software is freely available to academic users on www.jyu.fi/bioenv/en/divisions/smb/virus/primordialevo. Additional information is provided on the web page.

PrimordialEvo simulates the evolution and dynamics of an RNA-world-like gene community. In the system proto-cells represent abiotic compartments that can harbor genes but are themselves not capable of self-replicating. The genes of PrimordialEvo are ribozyme-like agents and thus competent for promoting simulated enzymatic tasks and serving as templates for their replication. The simulation is applied through rules which are iterated for the system to mimic the events of a hypothetical primordial communal environment. However, it must be noted that, unlike in many previous studies, PrimordialEvo does not take into account, for example, the length of replicated molecules, their exact sequences, three dimensional shapes, locations of individual domains in genes, temperatures of different regions of the system, the chemistry of resources or the evolution of replication fidelities.

The structure and the acting components of the system are described below.

Rules

The rules of PrimordialEvo remain constant throughout the simulation and they are applied individually for each proto-cell in every simulation cycle. The cycles are the driving force of the simulation and during each cycle various events may occur. The basic set of rules themselves is fixed (e.g. the order of variation operators), but most of the actual parameters associated with each rule can be adjusted by the user (e.g. the probability for a mutation to occur).

The rules are:

  1. 1.

    Replication of genes. These rules attempt to replicate the genes within the proto-cell. The amount of actual replications to take place is dependent on available resources and on the number of replicators. This step also attempts to destroy replication parasites (see section”Parasites”) by replication parasite destroyers if the requirements for resources are met.

  2. 2.

    Replication of selfish viruses. This step attempts to replicate the selfish, resource consuming viruses that only replicate themselves. The amount of available resources may limit their replication.

  3. 3.

    Formation of vesicles. Vesicles are formed according to system settings and the amount of budders within the proto-cell. These vesicles randomly fuse with a proto-cell that is located close to the original proto-cell.

  4. 4.

    Formation of selfish viruses. Selfish virus genes can induce the formation of vesicles (that enclose only the virus genes). Similar to the above, these vesicles fuse with nearby proto-cells and thus introduce virus genes into the proto-cells.

  5. 5.

    Collection of resources. Proto-cells are given resources according to the number of resource collectors they harbor and according to the pre-defined amount of passive inflow of resources.

  6. 6.

    Degradation of genes. This rule removes degraded genes from proto-cells.

  7. 7.

    Decomposition and formation of cell walls. Cell walls generating genes gradually synthesize molecules that build a cell wall around the proto-cell. Some of the cell wall (up to 3 %) is decomposed in every cycle.

  8. 8.

    Silencing of selfish viruses. Selfish virus silencers attempt to silence, i.e. destroy, selfish virus genes.

Proto-cell Community (Compartment Matrix)

The proto-cell community is constructed out of cells that form a fixed, three dimensional cube (Fig. 1a) that simulates primordial, inorganic compartments. Genes can exist only inside the proto-cells.

Fig. 1
figure 1

a Depiction of the proto-cell matrix and an individual cell. b Representation of the pathways that a gene may take as it is being replicated by a replicator

Full size image

Proto-cells

In PrimordialEvo proto-cells are compartments in which genes reside. The proto-cells are surrounded by a virtual membrane. Vesicles can bud from or fuse with the proto-cells. Proto-cells have a user-defined capacity for the amount of genes throughout the simulation. Proto-cells can have an organic cell wall which is encoded by cell wall generating genes. The cell wall coverage inhibits the vesicle budding and vesicle fusing with a probability equal to the percentage of cell wall cover. For example, if the proto-cell is surrounded by a cell wall for 70 % of its surface area, then an incoming vesicle has only a 30 % chance to fuse with this proto-cell. The cell wall deteriorates constantly, thus the cell wall generating genes are needed for maintaining the wall. Three different component species form the actual wall. They are all required for the wall to fully surround the proto-cell as one component can cover one third of the proto-cell surface. Cell walls may be disabled from the simulation setup. Proto-cells host five different resources which are exploited by the genes in order to facilitate gene replication, cell wall formation or defensive measures.

Genes

The genes of PrimordialEvo are capable of catalyzing certain events and serve as templates for their own replication. All genes deteriorate at a rate defined by the user.

Gene Replication

The genes are replicated only through the assistance of a replicator. However, the resource consuming, self-replicating parasites, i.e. the selfish viruses, can replicate without the help of other replicators. There is an equal chance for any gene in the proto-cell to become replicated by the replicators (excluding replication parasites, see Parasites). Replication occurs only if there are enough resources available to support the event. A single replicator can replicate one gene within a simulation cycle.

Mutations

Upon the replication of a gene, there is a user-defined chance for the gene to become mutated. Figure 1b describes the different pathways that a replicated gene may take. Mutation transforms a gene to perform another task or mutation can lead to a loss of function. If the mutation does not result in a loss of function, the gene is randomly mutated into any other gene type (excluding replication parasites and selfish viruses, for which the probability can be set individually). However, the regular mutations are calculated first and only if no mutations occur the gene can turn into one of the parasite-types. Moreover, only budders and replicators can mutate into selfish viruses as both of these gene-types represent a logical precursor of a selfish virus.

Vesicle Budding

Genes may be transferred from one proto-cell to another within vesicles. The first method for vesicle formation is an induced budding which is catalyzed by budder-genes. The user can define the required amount of budder-genes for vesicle formation to occur. The vesicles enclose a user-defined amount of genes, and from this amount the number of required budders is reduced. The second method is a random budding which occurs for each proto-cell at a chance defined by the user. The randomly budded vesicle encloses an amount of genes equal to the free gene slots in induced vesicles. The third method is the budding of selfish viruses as they catalyze their own budding. Only selfish virus genes are carried within these vesicles. While vesicles and proto-cells have sometimes been used as synonyms in literature, they should not to be confused here for one another as vesicles are only the carriers of genetic information between proto-cells.

Parasites

In PrimordialEvo there are two types of parasites: replication parasites and selfish viruses. The function of replication parasites is based on the original computational work of Manfred Eigen and Peter Schuster in 1979 with their hypercycle model and on parasites that may exploit these cycles (Maynard Smith 1979; Bresch et al. 1980). Replication parasites are good templates for replication, but they do not catalyze any reactions themselves. If replication of a gene occurs in a proto-cell, there is a user-defined chance for a replication parasite to be replicated instead of any other gene in the proto-cell. The expected amount of replication parasite replications per cycle is the number replicators multiplied with the user defined chance and the number of replication parasites. Replication parasites may arise when a gene is replicated.

The second parasite type is a selfish virus. These viruses arise from budder-genes or from replicators through special, user-defined mutations. Selfish viruses are able to replicate themselves and induce their own budding out of the mother proto-cell. The replication of a selfish virus-gene uses the same amount of resources as the replication of other genes. The possibility for a selfish virus to replicate itself can be defined by the user. The probability for a selfish virus to form a vesicle is the same as the probability for induced budding. All parasites and their anti-parasites may be disabled in the simulation setup.

It would be natural to assume that parasites may evolve so that they are no longer recognized by anti-parasites and similarly that anti-parasites may evolve yet again to recognize parasites. However, we deliberately left out from the simulator this potential for evolutionary arms-race between parasites and their anti-parasites.

Anti-parasites

There are two types of anti-parasite genes in PrimordialEvo: the replication parasite destroyers and the selfish virus silencers. As the name indicates, replication parasite destroyers can destroy replication parasites and selfish virus silencers can destroy selfish viruses. The chance for the silencers and the destroyers to attempt a destructive measure can be defined by the user. Any successful attempt costs a user-defined amount of resources.

Resources

There are five different numerically simulated resources in PrimordialEvo. The user can define the following values for the system: the initial resources in the proto-cells, the maximum resources a proto-cell can hold, the passive inflow of resources into the proto-cells and the amount of resources each resource-collector acquires per cycle. The passive inflow of resources indicates the amount of resources that are automatically added to each proto-cell per cycle. Resources are used to replicate genes, to perform destructive measures against parasites and to build cell walls.

Simulation Setups

There is a variety of system parameters that can be defined prior to the execution of the simulation (as noted above). The parameters and their values used in the experiments are listed in Table 1.

Table 1 Adjustable attributes of the simulator and their values in the present study
Full size table

Results and Discussion

Events During a Typical Simulation

The dynamics and events of a laterally evolving community of ribozyme-like agents were dissected and analyzed to study general attributes of the presented form of simulated evolution. Many conditions were roughly screened for the changes in the amounts of different genes and resources. Long term behaviors of various systems were observed and the equilibriums were analyzed. Generally equilibriums of systems with high amount of genes are stable and relatively constant amounts of different genes are maintained. More delicate systems (e.g. those that have high gene degradation rates and low gene capacities) are prone to oscillate between exponentially growing and degrading phases. A typical simulation, in which clear evolution takes place, was selected for closer examination for the events that take place after the first replicators ignite the life in the compartment matrix. The exact values of this simulation setup are listed in Table 1. The simulation was repeated ten times in order to rule out stochastic randomness from the observations. The time dynamics of various attributes of the simulated system are presented in Fig. 2. At first, there is the spread of genes throughout the system within the vesicles. Spread is followed by the takeover of the proto-cells by replication parasites which are then dominated by anti-parasites (as they emerge). Eventually all the needed resource-collecting genes appear, thus hastening the resource gathering of proto-cells. The reduction of the resource limitations on replication causes the exponential growth of genes in the system and finally the system reaches its maximum of genes. However, more and more genes start losing their function because of the loss-of-function mutations (and replicators further amplify the useless genes) and thus the amount of resource collecting genes, replicators and such decreases. Some compartments might lose them even completely. This disappearance is for two reasons: firstly because of random horizontal movement of resource collectors and/or replicators to other compartments within vesicles, and secondly because replicators select the genes for replication randomly — thus making it possible for critical genes to become accidentally completely degraded. Initially, the disappearing of necessary and useful genes such as resource collectors does not seem to affect the total gene count of the system since there is a surplus of, e.g., resource-collectors anyway. As the number of cycles increases, the loss of resource collectors begins to limit the replication rate once again in some of the compartments and the total number of genes in the system starts to decrease. However, the lost genes can be replaced either by mutations that again come up with the innovation of resource collectors or due to horizontal movement of the innovations from neighboring compartments. In the long term experiments we observed that low gene maximum and fast gene degradation makes the compartments and the whole system more easily to fall into this collapsing sequence.

Fig. 2
figure 2

Events during a simulation. a Initial burst of induced vesicle formations spreads the replicator and budder genes to fill the system. b Replication parasites take over some of the cells. c Anti-parasites begin to destroy the replication parasites. d Different resource collecting genes emerge and thus free the replicators to replicate genes without resource restrictions. e Resource collectors allow the gene-count of the system to grow exponentially. f Useless genes begin to accumulate into the cells due to the mutations causing the loss of function. g Accumulation of useless genes begins to limit the replication of genes in some of the proto-cells. X-axis: number of cycles; Y-axis: the number of particular genes (b,c,e and f), the amount of induced vesicle formations (a) and the maximum potential amount of resources that can be collected (d)

Full size image

Natural Selection of Resource Collectors and Anti-parasites within a Laterally Evolving Community of Ribozyme-like Agents

Given that a laterally evolving community does not consist of the traditional, genetically established species and genomes that could be either favored or disfavored by natural selection, it would be of importance to demonstrate if natural selection drives the selection of individual genes even among a set of loosely bound genetic agents within the simulated compartment matrix. Previously it has been observed that the gene-centered evolution (i.e. the “selfish-gene” point of view) is an ideal approach for understanding the biological world (see e.g. Dawkins 1976). The gene-centered evolution most probably emerged at the time when the first genetically encoded replicating agents appeared on Earth. It has been demonstrated that the single stranded RNA genome of bacteriophage Qβ evolves in vitro when the molecule is repeatedly replicated by RNA-replicases (Kacian et al. 1972). Such enzymatic RNA replication has also been situated into self-replicating vesicles to model minimal cells (Oberholzer et al. 1995). Recently Lincoln and Joyce were able to produce a self-sustaining replication system consisting of only RNA-enzymes (Lincoln and Joyce 2009). Moreover, in vitro evolution of molecular cooperation (Ellinger et al. 1998) and artificial selection of novel catalytic RNA functions have been experimentally tested (Curtis and Bartel 2005). We have here studied how a community of such RNA-enzymes may behave after the emergence of first self-replicating molecule and how natural selection can favor some functions in the laterally evolving system.

As seen in Fig. 2, resource collectors are amplified in the system according to the demand of the particular resource for gene replication (at least when all non-parasitic genes are replicated at equal chance). The selection of required resource collectors can be explained by group selection (Wilson 1975), which has often been applied in studies concerning primordial forms of life (see e.g. Takeuchi and Hogeweg 2008; Könnyu et al 2008). If genes that are able to improve the compartments’ gene replication rates moved to a new compartment, then the new compartment is more competent of reproducing vesicles, which would allow this compartment to yet again distribute its genetic content to neighboring proto-cells. Such selection has been hypothesized to also favor genetic functions that facilitate the movement of genetic material between compartments (Jalasvuori and Bamford 2008), such as the vesicle formation inducing genes (as observed in Fig. 2). Moreover and in support for group selection, there is no amplification of genes above the levels of others in systems consisting of only one single compartment.

In some long term simulations it was observed that the initially amplified resource collectors eventually decrease to similar levels with collectors for which no selection is present (Fig. 3a). However, selection has not disappeared in these situations but instead these well established systems can maintain high number of genes regardless of their favorability that the resource collectors (at their levels in the equilibrium) are able to saturate the compartment with resources during each simulation cycle. When the long term behavior of a more delicate system (which has relatively low gene maximum and number of compartments) is being studied, then differences in the amounts of resource collecting genes can be observed throughout the simulation (Fig. 4). However, such systems are prone to entirely lose critically important genes and may have periods during which life only barely survives. Nevertheless, these delicate systems roughly represent systems that most of the time contain free uninhabited compartments and thus group selection of beneficial resource collectors can continue to amplify useful genes indefinitely. If there was a system with a very high gene maximum and the number of compartments was huge, then the exponential growth phase would continue for long periods of time and the differences in the amounts of replication potential improving genes and other genes would become enormous.

Fig. 3
figure 3

Testing the selection of resource collecting genes and anti-parasite genes in long term (100000 cycle) simulations. a The maximum potential amount of resources collected by resource collecting genes during a simulation in which parasites were allowed to emerge. b Time dynamics of anti-parasite genes in presence and absence of parasites. c Time dynamics of replication parasites during the same simulation with A and B. X-axis: number of cycles

Full size image
Fig. 4
figure 4

A simulation of a delicate system for 100000 cycles (see Table 1 for details). Resource one (blue) was the most required resource for replication and it was observed to be the highest most of the time throughout the simulation. X-axis: number of cycles

Full size image

However, in order to test the selection of genes that are not straightforwardly promoting the collection of limiting resources, we tested whether enabling replication parasites amplifies the number of replication parasite destroyers (i.e. anti-parasites). The potential selection of anti-parasites would intuitively occur due to the removal of harmful parasites (that hinder the replication of other agents) from the system.

In the simulation systems that allowed the emergence of replication parasites, the number of replication parasite destroyers were almost exclusively elevated above the levels of the control simulations during the exponential growth phase (Fig. 3b). This clearly suggests that anti-parasites are favored in systems where parasites may appear. However, later on the number of anti-parasites decrease to equal levels with those genes for which no selection exists. Should all the anti-parasites become eradicated from the compartment, then selection can be observed again. This is because only few anti-parasites are needed to promptly destroy any emerging parasites. If no anti-parasites are present, then parasites can conquer the system (as seen in Fig. 3c). In such a case the emergence of anti-parasites form the prerequisite for the system to re-establish the replication of non-parasitic genes and the exponential growth phase (as seen in Fig. 3b).

Previously, the presence of parasites has been observed to cause the evolution of increased complexity in Monte-Carlo simulated RNA-replicator communities (Takeuchi and Hogeweg 2008) and in experimental evolution systems consisting of ribozymes (Hanczyc and Dorit 1998). We suggest that, while parasites seem to be responsible for increasing complexity in these systems (partly) because they produce raw material for evolution of novel functions, they also promote higher genetic complexity by favoring certain otherwise useless genes (as anti-parasites).

Favorability of Induced Vesicle Formation within the Community

It has been suggested that genes inducing the vesicle formation within a community of proto-cells should promote the exchange of innovations between proto-cells as these vesicles can act as carriers of useful genes (Jalasvuori and Bamford 2008). Moreover, it was proposed that, if the very first membrane bound replicators produce a mutant gene that is able to induce vesicle budding, then the vesicles would greatly accelerate the rate at which life spreads to fill up the system.

A system of hundred and twenty five proto-cells was simulated for twenty thousand cycles in three different system setups (for setup details, see Table 1). The first simulation setup allowed budder-genes to induce vesicle formation but the random formation of vesicles was set to be a rare event (0.1 % chance per proto-cell per cycle). The first setup is the same as the one that was analyzed in section “Events during a typical simulation”. In the second and third simulation setups the induced vesicle formation was suppressed. However, in the third simulation the random formation of vesicles was set to 5 % per proto-cell per cycle, which thus roughly represented a compartment matrix with freely diffusing genes inside. All simulation setups were executed sufficient times in order to recognize patterns in the stochastic form of evolution (n = 10 was used as the statistics were generally similar with respective setups).

The systems begun to evolve from ten genes (replicators and, in setup one, budders) that were situated into a single proto-cell (at position [x = 0, y = 0, z = 0] of the three dimensional matrix of 125 proto-cells). The maximum amount of genes varied greatly between systems as the number of cycles advanced. The system with attributes allowing only rare vesicle formation (i.e. the second setup) was never able to evolve a gene-community of more than three thousand genes while the other two systems nearly always (excluding one of the ten simulations of the first setup in which the gene count was at its maximum hardly more than five thousand) reached a gene-count of at least hundred and sixty thousand. The evolution of the amount of genes in a typical, representative simulation of each three setups is presented in Fig. 5a.

Fig. 5
figure 5

a Illustration of the changes in the abundance of genes and b the changes in the potential amount of resources that can be collected by resource-collectors in a representative simulation of each three setups (see text). Requirement for gene replication was six, four and three times more dependant on resource 1 (blue), resource 2 (pink) and resource 3 (yellow), respectively, than on resource 4 (cyan). Resource 5 was not required for replication (purple). X-axis: number of cycles; Y-axis: the number of genes (a) or the maximum collection potential of each resource (b)

Full size image

Interestingly, selection towards favoring the most needed resource collecting genes occurred in the first and the third systems while being absent in the second system (Fig. 5b). Isolated compartments, even thought they may sometimes exchange genetic material with neighboring proto-cells, seem to be unable to come up with all the required evolutionary innovations to improve the gene replication rates (at least within the timescales of the simulations performed here). In other words, when multiple innovations are required, the first innovation that might evolve in a single compartment can become degraded before the emergence of second useful one. Thus each compartment relays mostly on passive inflow of resources, waiting for the unlike event of rapid appearance of multiple innovations. This suggests (and supports the previously made hypothesis) that an active exchange of innovations between proto-cells is promoting the evolution of genetic complexity as innovations that emerge in any given compartment is not the restricted property of that proto-cell alone. Moreover, the rate at which genes increased in number in the first setup was more rapid than in the third setup with passive transfer of genes. As was hypothesized before, gene-encoded mechanism that allows compartments to exchange information seem to be more effective and could therefore outcompete systems in which genes only randomly move between compartments. However, our observations also suggest that systems with high rates of inter-compartment genetic exchange are required to have the potential for anti-parasite innovations in order to survive long periods of time. In absence of anti-parasites the uncontrollably amplified parasites can move along with the vesicles and destroy life in the rapidly spreading epidemic. This is in concert with previous studies that have suggested compartmentalization and relatively rare horizontal gene transfer to protect against parasites as in these systems parasites destroy themselves before they spread to neighboring proto-cells.

The ratio of useful genes was observed to be the highest in Setup 2. This is merely because the compartments have low amount of resources and therefore they simply cannot support many useless genes. If the compartments of system Setup 2 had evolved high levels of useless genes, they would have self selected themselves out as replicators used majority of the resources for replicating useless genes. As the final replicator becomes degraded due to lack of resources, there is nothing to support the replication of useless genes, thus leading eventually into an empty compartment. At some point, horizontal movement of a new replicator would restart life there, again selecting for low amounts of useless genes.

Conclusions

In this paper we present a rule-based computation system, PrimordialEvo, for simulating a community of ribozyme-like agents. These ribozyme-genes may perform enzymatic tasks, like replicate and produce resources, and also serve as templates for their own replication. The community is situated in a three dimensional matrix of (inorganic) compartments. Gene flux between compartments can be set to be either a random or gene-induced event.

PrimordialEvo was used to analyze the evolutionary dynamics and some general features of a simulated community of genes. Proto-cells that have high gene capacity were noted to be more likely to establish stable systems whereas proto-cells that have low gene capacity and quickly degrading genes were unstable and prone to collapse. This could imply that first ribozyme-like agents were required to be able to replicate into sufficient levels for life to survive. The importance of innovation sharing for early evolution of life has been convincingly argued by Vetsigian and colleagues in 2006. We observed that innovation-sharing between compartments accelerates the overall rate of replication success within the matrix, which thus supports these previously made arguments. However, the innovation sharing systems are required to be able to come up with the innovation of anti-parasites as otherwise parasite epidemics destroy these systems. The simulated system was also demonstrated to increase the frequency of genes according to their ability to support the replication of genes within the compartment. Moreover, the presence of replication parasites lead to amplified levels of anti-parasites. This was observed to be due to the anti-parasites ability to clear the system from exploiters that hindered the replication efficiency of replicators. The selection of genes was observed to be present only when the system consisted of multiple compartments, therefore suggesting that group selection was promoting the amplification of certain genes above others.