Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Impact of composition on the dynamics of autocatalytic sets

Impact of composition on the dynamics of autocatalytic sets Alessandro Ravoni Department of Mathematics and Physics, University of Roma Tre, Via della Vasca Navale 84, 00146 Rome, Italy Autocatalytic sets are sets of entities that mutually catalyse each other's production through chemical reactions from a basic food source. Recently, the re exively autocat- alytic and food generated theory has introduced a formal de nition of autocatalytic sets which has provided promising results in the context of the origin of life. However, the link between the structure of autocatalytic sets and the possibility of di erent long-term behaviours is still unclear. In this work, we study how di erent interactions among autocatalytic sets a ect the emergent dynamics. To this aim, we develop a model in which interactions are presented through composition operations among networks, and the dynamics of the networks is reproduced via stochastic simulations. We nd that the dynamical emergence of the autocatalytic sets depends on the adopted composition operations. In particular, operations involving entities that are sources for autocatalytic sets can promote the formation of di erent autocatalytic subsets, opening the door to various long-term behaviours. Keywords: Autocatalytic sets; Origin of life; Network Composition; Stochastic Petri nets; Binary polymer model I. INTRODUCTION Moreover, RAF theory has successfully proved that RAF sets have a hierarchical structure, where the largest set of reactions within a CRS satisfying RAF prop- The exact sequence of events that led to the formation erty (the so-called maxRAF) has many subsets that of the rst living organisms from non-living matter is are smaller RAF sets themselves (Hordijk et al., 2012; still a topic under debate (Benner et al., 2012; Bernhardt, Hordijk and Steel, 2014, 2018). The latter is a pecu- 2012; Luisi, 2016; Rasmussen et al., 2004; Szostak, 2017). liarity that makes RAF sets potentially suitable for ex- On the other hand, some particular qualities of early periencing adaptive evolution (a feature that is gener- organisms are commonly accepted and are well known. ally referred to as evolvability); i.e., to collect evolution- Among these, we can point out the self-replication ability ary changes bene cial for survival and reproduction in a of the early life forms (Higgs and Lehman, 2015; Kau - given environment. Indeed, it has been argued that the man, 1993; Luisi, 2016; Nghe et al., 2015; Rasmussen et autocatalytic subsets present within the structure of the al., 2016). In this scenario, the autocatalytic sets (ASs) maxRAF could be the elementary units on which natu- are of great interest. Introduced by Kau man (Kau - ral selection can act (Hordijk and Steel, 2014; Hordijk et man, 1971, 1986, 1993), ASs are sets of entities capable al., 2018b; Vasas et al., 2012): the availability of sponta- of spontaneous emergence and self-reproduction through neous reactions would allow the occurrence of mutations catalytic reactions, starting from a nite set of entities and, consequently, the appearance of novel autocatalytic assumed to be available in the environment. There are subsets able to replicate themselves with di erent rates several de nitions of ASs in the literature (see, for in- and competing with each other. stance, (Jain and Krishna, 1998; Sharov, 1991)). Re- cently, Hordijk et al. (2011) have introduced the notion of However, rst results show that the asymptotic dy- re exively autocatalytic and food generated (RAF) sets, namics of simple RAF sets eventually reaches the state a formal de nition of ASs in the framework of chemi- in which all the reactions of the maxRAF occur catalyt- cal reaction systems (CRSs) (see Section II for de ni- ically (Hordijk et al., 2018b; Vasas et al., 2012). In this tions). Properties of RAF sets have been studied by var- state, all the elementary autocatalytic units coexist with- ious authors, and among the most important results (see out e ectively competing with each other, thus leaving (Hordijk and Steel, 2017, 2018) for more details) note the no room for adaptive evolution (Hordijk et al., 2018b; implementation of a polynomial time algorithm able to Vasas et al., 2010, 2012). The evolvability of RAF sets identify the presence of a RAF set in a general network can be restored by embedding them into compartments of interacting entities (Horidjk and Steel, 2004; Hordijk and allowing the sharing of resources and the exchange et al., 2011; Hordijk and Steel, 2012) and the detection of chemical molecules (Hordijk et al., 2018b; Kau man, of an autocatalytic structure in the metabolic network of 2011; Serra and Villani, 2019; Vasas et al., 2012). In fact, Escherichia Coli (Sousa et al., 2015) and ancient anaero- through numerical simulations, it has been observed that bic autotrophs (Xavier et al., 2020). RAF sets enclosed in semipermeable protocells can reach arXiv:2009.05958v1 [nlin.AO] 13 Sep 2020 2 di erent asymptotic states (Serra and Villani, 2019), and { ; is a pseudo-entity representing the environment that spatially separated RAF sets consuming the same and s 2 S is a food entity if there exists a reaction food source can give rise to di erent combinations of r 2 R such that (;; r) 2 E and (r; s) 2 E . r r competing autocatalytic subsets (Hordijk et al., 2018b), suggesting that the evolvability of RAF sets is related to Note that E is called the spontaneous reactions set. the interactions among RAF sets themselves. The edges (s; r) 2 E can be interpreted as source entity In this work we investigate this latter point. In par- s is consumed by reaction r, while in edges (r; s) 2 E ticular, we study the role of various interactions among entity s is produced by reaction r. i denotes a reaction RAF sets in order to understand how these interactions such that (;; i ); (i ; s) 2 E , implying an input reaction s s r a ect the emergent dynamics. To this aim, we use the producing a food entity. Moreover, we introduce out ow stochastic Petri nets (SPNs) formalism (Haas, 2006; Mol- reactions o such that (s; o ); (o ;;) 2 E . Thus, the s s s r loy, 1982) to represent and evolve RAF sets. Further- CRS is a ow reactor that allows in ow and out ow of more, we introduce some composition operations acting entities from and towards the environment. on nets, which correspond to di erent interactions among Let R represent a subset of R. The closure cl (F ) RAF sets. In this framework, assuming that the entire is de ned to be the (unique) minimal subset of S that maxRAF set always emerges in an isolated RAF set, contains F together with all entities that can be produced our goal is to nd some composition operations under from F by repeated applications of reactions in R. Note which the dynamical appearance of the maxRAF is not that cl (F ) is well de ned and nite (Horidjk and Steel, invariant. This means that the corresponding interac- 2004). Given a CRS (S; R; C; F ), a RAF set is a set of tion causes only some of the maxRAF subsets to emerge, reactions R  R (and associated entities) that satis es allowing the existence of multiple long-term behaviours the following properties: required for the evolvability of RAF sets. The paper is organised as follows. In Section II we in- 0 { Re exively autocatalytic: for each reaction r 2 R troduce the de nitions of RAF sets and SPNs. In Section there exists at least one entity s 2 cl (F ) such III we describe the model we use to evolve nets and we that (s; r) 2 C ; introduce the composition operations. In Section IV we present and analyse the results obtained by simulating { F -generated: for each reaction r 2 R and for each the dynamics of various composed RAF sets. Finally, in entity s 2 S such that (s; r) 2 E , it is s 2 cl (F ). r R Section V we discuss and demonstrate the conclusions. Thus, a RAF set is a set of reactions able to catalytically produce all its source entities starting from a suitable II. BACKGROUND food set. It is also possible to de ne the closure of a set A. Re exively autocatalytic and food generated sets of reactions, introducing the notion of closed RAF sets (Smith et al., 2014). Given a CRS (S; R; C; F ), a subset R of R is said to be a closed RAF set if: In RAF theory (Horidjk and Steel, 2004; Hordijk et al., 2011), a network of interacting entities is represented by a CRS. Following previous de nitions (Horidjk and { R is a RAF set; Steel, 2004; Hordijk and Steel, 2017; Lohn et al., 1998), we introduce a CRS as a tuple (S; R; C ) such that: { 8 r such that all its source entities and at least one catalyst are either part of the set F or are produced { S is a set of entities; 0 0 by a reaction from R , it is r 2 R . { R is a set of reactions,  ! , where ;  2 S are the reactants and products of a reaction, respec- Authors of (Hordijk et al., 2018a) describe a procedure tively; to detect closed RAF sets in a generic CRS. It has been argued that closed RAF sets are associated with { C is a catalysis set, that is, a set of pairs f(s; r); s 2 the attractors of the dynamics of a CRS and there- S ; r 2 Rg indicating the entity s as the catalyst of fore represent the relevant units capable of experienc- reaction r. ing adaptive evolution (Hordijk et al., 2018a,b; Smith et al., 2014; Vasas et al., 2012). FIG. 1 shows an exam- We also de ne a food set F  S such that entities s 2 F ple of a RAF set and its constituent closed RAF sets. are assumed to be available from the environment. We We exemplify our point by introducing a set of entities describe a CRS as a bipartite graph such that: S = fA; B; C; D; E; F; G; H; I; L; M; N; O; Pg and a food set F = fA; B; D; F; I; Mg. In this example, the RAF set { nodes are of two kinds V = S [ R; is composed of the following reactions (an entity above { edges are of two kinds E = E [ C ; the arrow of a reaction indicates the catalyst associated r 3 B. Stochastic Petri nets A Petri net consists of (see Petri and Reisig (Petri and Reisig, 2008) for further details): { a nite set of place P ; { a nite set of transitions W ; { functions b; e : P  W ! N. Here b(p; w) and e(p; w) are the number of edges from place p to transition w and from transition w to place p, respectively. The sets b(w)  P and e(w)  P are the sets of places connected to transition w by at least one edge. A marking X of a Petri net is a map X : P ! N that assigns a number of tokens to each place. In fact, a FIG. 1 RAF set example. Entities are represented by let- marking X identi es a state of the system in the space of ters (dark-green for food-entities, black for non-food entities). possible con gurations of tokens available in each place. The food set is surrounded by a dark-green dashed rectangle. With x , we indicate the number of tokens of place p Reactions are displayed by coloured squares. A black arrow available in marking X . Firing a transition w consumes emerging from a letter towards a square (from a square point- b(p; w) tokens from each of its input places p 2 b(w), ing at a letter) indicates the corresponding entity is a source (a product) for that reaction. Red dashed arrows indicate and produces e(p ; w) tokens in each of its output places catalysis. The maxRAF set consists of three closed RAF sets: p 2 e(w). For each marking X , a transition w is enabled (1) (2) R = fr ; r g (green squares), R = fr ; r ; r ; r g (ma- 1 3 1 3 5 8 (it may re) if there are enough tokens in its input places (2) genta and green squares), R = fr ; r ; r ; r ; r ; r g (blue 1 3 2 4 6 7 making the consumption possible. This shall occur, if and green squares). and only if X (p)  b(p; w), 8p 2 P . A stochastic Petri net (Haas, 2006; Molloy, 1982) is a Petri net for which each transition is equipped with a (possibly marking de- with that reaction): pendent) rate for the exponentially distributed transition ring times.  denotes the set of ring rates of a SPN. r : A + B ! C Note that the evolution of a SPN with exponentially dis- tributed transition rates is isomorphic to continuous-time r : A + D ! E Markov chain (Molloy, 1982). Petri net formalism provides a suitable environment r : C + F ! G for studying the composition of networks, with both a r : B + E ! H 4 computational and theoretical approach; the latter, in P particular, in the context of category theory (Baez and r : C + I ! L Pollard, 2017). Moreover, this formalism can describe r : C + M ! N nets with di erent dynamics (Vazquez and Silva, 2011). The authors will investigate these aspects in a forthcom- r : D + M ! O ing work. r : F + L ! P III. THE MODEL The maxRAF set consists of three closed RAF sets: A. Building the net (1) (2) (2) R = fr ; r g, R = fr ; r ; r ; r g, R = 1 3 1 3 5 8 fr ; r ; r ; r ; r ; r g. Note that the closed RAF set Given a CRS (S; R; C; F ) (or a set of reactions and 1 3 2 4 6 7 (1) R = fr ; r g is also a constructively autocatalytic and associated entities that is a RAF set), we build a SPN by 1 3 food generated (CAF) set (Mossel and Steel, 2005); i.e. adding a place p for each species s 2 S and a transition w a RAF set for which reactions can take place on the con- for each reaction r 2 R such that b(w) = (r) and e(w) = dition that their catalysts are already made available by (r), where (r) and (r) are the set of all sources and the occurrence of catalysed reactions (starting from F ). targets entities of edges (s; r); (r; s) 2 E , respectively. More restrictive conditions of this kind result in higher Note that with this notation we consider both in owing rates of catalysis that form a CAF set (Mossel and Steel, and out owing transitions. Moreover, for each catalysis 2005). All the RAF sets studied in this work contain one (s; r) 2 C we add a transition w such that b(w) = (r)[s RAF subset which in turn is a CAF set. and e(w) = (r) [ s. Hereafter, we use S and R to 4 indicate both species and reactions of a CRS and places work (Feinberg, 1995). The authors are currently inves- and transitions of a SPN. The rates (r) associated with tigating this aspect in an upcoming work. each transition r are marking dependent rates: With this limitation, all the spontaneous reactions of our model are binary reactions (i.e., reactions with two (r) = h (X ) : (1) r r reactants, possibly of the same entity type), while all the catalysed reactions are ternary reactions. Even if Here,  is a xed constant depending on the type of its ternary reactions are rare, they can represent a rst ap- corresponding reaction in the CRS ( = f ;  ;  ;  g r s c i o proximation of two or more elemental reactions, such as in which  speci es spontaneous, catalysed, in owing the sequence of reactions of an enzyme catalysis (Gille- and out owing reactions, respectively) and h (M ) is a spie, 2007). value proportional to the number of combinations of to- kens available in the input places of transition r in the state X . Thus, explicitly, we shall have: B. Composition b(j; r) j j h (X ) = ; (2) We model interactions between CRSs as composition jb(r)j1 b(j; r) operations between SPNs. First, note that RAF sets sat- isfy the following conditions (hereafter, we refer to these where the product is among all the input places of tran- as the inclusion conditions) (Horidjk and Steel, 2004): sition r. We set functions b and e such that the in owing transitions do not consume tokens of the pseudo-entity if R is RAF in (S ; R ; C ; F ), it is RAF also 1 1 1 1 ; and produce a xed value of tokens of the food enti- in (S ; R ; C ; F ), if conditions S  S ; R 2 2 2 2 1 2 1 ties, while the out owing transitions consume a token of R ; C  C ; F  F are satis ed; 2 1 2 1 2 the out owing entities and do not produce tokens of the 0 0 if R is RAF (S ; R ; C ; F ) and R is RAF in 1 1 1 1 1 2 pseudo-entity ;. Thus, the rate of out owing transitions 0 0 (S ; R ; C ; F ), then R [ R is RAF in (S [ 2 2 2 2 1 is proportional to the amount of tokens of the out owing 1 2 S ; R [ R ; C [ C ; F [ F ). 2 1 2 1 2 1 2 entity, while the rate of in owing transitions is indepen- dent of the state of the system. It is noteworthy that Thus, if composition does not remove entities from the the in owing of food elements still remains a stochastic food set or reactions belonging to a RAF set, it will not event. We add in owing transitions for entities not be- have any impact on its RAF property. However, the dy- longing to the originary food set F (setting the rate of namical behaviour of the composed system can be, gen- such transition equal to zero), if required for the purpose erally, di erent from that of the starting one. of composition between nets (see Section III.B). The dy- Let (S ; R ; b ; e ;  ) and (S ; R ; b ; e ;  ) be two 1 1 1 1 1 2 2 2 2 2 namics of the obtained SPN is described by the stochastic SPNs and let I; O be subsets of their in owing and out- mass action kinetics, that is the classical dynamics used owing transitions sets. Let be the equivalence relation to represent chemical reactions, assuming a well-stirred such that s  s if i 2 I and o 2 O for some choice of s s system (Anderson and Kurtz, 2011). I; O. S denotes the set of places identi ed by relation The CRSs used in this work are generated according . We de ne the following composition operations: to the binary polymer model (BPM) (Kau man, 1986). The BPM produces a CRS where the entities set S con- CO : (S ; S ) ! S = S [ S ; I 1 2  1 2 sists of all bit strings up to (and including) a maximum (R ; R ) ! R = (R [ R [ R )n (O [ I ); 1 2  1 2 I 1 2 length N , and the reaction set R consists of condensation (b ; b ) ! b = b [ b [ b ; 1 2  1 2 I and cleavage reactions. Condensation reaction is a con- (e ; e ) ! e = e [ e [ e ; catenation of two bit strings resulting in a longer string, 1 2  1 2 I and cleavage reaction cats a bit string into two smaller ( ;  ) !  =  [  [  ; 1 2  1 2 I ones. The food set is represented by all entities with 0 0 R := frj b(r) = s; e(r) = s ; 8 s 2 S ; s 2 I 1 a length less than or equal to a xed length l (we set S such that s  s g; l = 2), and each entity can be a catalyst of each reaction := f(r)j (r) = h (X ) ; 8 r 2 R g. with a certain probability xed a priori. I r f I We chose to only allow the condensation reactions to CO : (S ; S ) ! S = S [ S ; II 1 2  1 2 occur in the system. Note that the technique we use is (R ; R ) ! R = (R [ R [ R )n (O [ I ); 1 2  1 2 II 1 2 also applicable in the case where cleavage reactions are allowed. In fact, none of the operations we introduce (b ; b ) ! b = b [ b [ b ; 1 2  1 2 II in Section III.B is a ected by the reversibility of the (e ; e ) ! e = e [ e [ e ; 1 2  1 2 II network reactions. Including cleavage reactions would ( ;  ) !  =  [  [  ; 1 2  1 2 II produce networks, in principle, with di erent dynamics, 0 00 00 since a network with irreversible reactions can have dif- R := frj b(r) = b(r ); e(r) = b(r ); 8 r 2 II 00 0 00 ferent topology with respect the equivalent reversible net- R such that b(r )  S and b(r )  b(r )g; 2  5 := f(r)j (r) = h (X ) ; 8 r 2 R g. It is worth to underline that, if there exists a s 2 S II r f II such that s 2 F , operations CO and CO can actu- I II CO : (S ; S ) ! S = fS t S g= ; ally modify the RAF property of net 2. In particu- III 1 2  1 2 lar, if transitions r 2 R are assumed to be spon- (I;II) (R ; R ) ! R = R [ R ; 1 2  1 2 taneous transitions ( <  ), net 2 could not be cat- f c (b ; b ) ! b = b [ b ; 1 2  1 2 alytically produced starting from the food set F . In this case, the composed net contains a RAF set R such (e ; e ) ! e = e [ e ; 1 2  1 2 0 0 that R  R  (R [ R ), with R = R if F  S . In- 1 1 2 1 ( ;  ) !  =  [  . 1 2  1 2 stead, if transitions r 2 R are assumed to be (auto) (I;II) catalysed transitions (   ), the whole composed net f c Here  is a constant value and (S ; R ; b ; e ;  ) is the shall be a RAF set. composed SPN. To summarise, all the composition operations we de- ne relate a set of places S that are input for out owing IV. RESULTS AND DISCUSSION transitions of a SPN, together with a set of places S that are input for in owing transitions of another SPN. The In this Section, we present the results regarding the formal addition of in owing transitions for places not be- impact of composition on the dynamics of RAF sets as longing to the food set enlarges the possible composition follows: we rst introduce the characteristics of simula- operations between SPNs. tions and the quantities taken into account. In Section Given a set S , operation CO adds a transition from I IV.A we present results for the non interacting nets, in S to S for each pair of places in S , while operation order to have a reference model for the interacting cases, CO adds a transition from S to S for each combina- II presented in Section IV.B. tion of places that appears as input of a transition in Starting from di erent instances of the BPM with the in owing net. Each combination corresponds to the N < 8, we use the RAF algorithm introduced in (Hordijk de nition of complex in the framework of chemical re- et al., 2011) in order to detect and select three di erent action networks (Feinberg, 1995). In fact, given a set of RAF sets, each of which contains more than one closed chemical species, a complex is de ned as a member of RAF set. Note that, even if di erent RAF sets have the the vector space generated by the species that provides same species set S, the set of reactions E will be di erent; the inputs (or the outputs) of a reaction (Anderson and i.e., various RAF sets have di erent chemistry. The RAF Kurtz, 2011; Feinberg, 1995). sets identi ed through this procedure constitute the col- Both operation CO and CO introduce a ux of en- I II lection on which we will carry out the study. Even if such tities from one net to another. The composite network a small collection cannot be taken as a solid statistical can therefore be seen as the union of two separate net- basis, it is still sucient for providing us with interest- works that evolve in parallel, communicating only with ing information. We duplicate each RAF set and let it (asymmetrical) exchange of chemical species. This could interacts with its copy by switching to representation of be, for instance, the case of two spatially separated pro- RAF set as an SPN and using one of the composition tocells, one of which can release molecules towards the operations introduced above. We simulated the dynam- other. According to this interpretation, the owing rate ics of the system using the standard Gillespie algorithm is a parameter that encompasses the characteristics (Gillespie, 1976, 1977), setting the volume of the system of the ow process (for example, cell permeability). Op- at V = 1 (arbitrary units). For each simulation, we per- eration CO , instead, merges each pair of places in S , form 100 independent runs of 10 time steps. One of III allowing transitions of the two original SPNs to operate the necessary conditions for a RAF set is the ability to on the glued set of places. Composing nets via opera- produce itself starting from the elements of the food set. tion CO actually produces a new single network. In Indeed, the initial state of the SPN is set such that: III this case, one can think of composite net as the result of mutations that enlarge a network (for instance, net x (t = 0) = x if s 2 F ; s 0 (3) 1), introducing new possible reactions and, consequently, x (t = 0) = 0 otherwise: new chemical species (corresponding to net 2). Here x is an arbitrary constant. The values of x ,  , 0 0 c and  are set such that the number of tokens of food i o places at t ! 1 is equal to 10 , for an SPN with in ow, Note that complexes play a major role in the framework of chem- out ow and all (and only) binary transitions having, as ical reaction networks theory. For instance, the de ciency theo- input, food places only (and ring rate  ). The value rems (Feinberg, 1995) are able to predict whether the dynamics of  is xed at  =  =10, while  varies such that of a large class of networks will have a stationary distribution, s s c f 1 6 starting from the topology of the reaction graph having com-  2 [ 10 ;  10 ]. It is noteworthy that the values f c c plexes as nodes. of these parameters are not taken from \in vivo" data, 6 but they have phenomenological motivations. Thus, al- though we can reasonably generalise the characteristics of the dynamics, quantities such as the species' produc- tion rate or the time evolution of the concentrations may di er from those in other similar stochastic simulations (Hordijk and Steel, 2012; Hordijk et al., 2018b). We focus our attention on the e ective appearance of a maxRAF set during the evolution of the system. In particular, we introduce the following quantities: 1) M (t) = , 8s 2= F ; 1+x (t) 2)  = minftj n(r)  i, 8r 2 Rg. Here n(r) is the number of executions of transition r, and R is the maxRAF set. Note that the natural con- dition x (t)  0 implies that each term contributing to FIG. 2 Isolated RAF set. Number of tokens over time of the computation of M (t) can possess, at most, the value non-food entities obtained in a simulation run of the isolated one. Both M (t) and  are calculated for each single s i RAF set shown in FIG. 1. After a transient time of t net that forms the composed net. Let m = M (t) be M s 0:006 (arbitrary units) all the entities are present in a large the mean value of M (t) for t ! 1. If all the non-food amount, with the exception of the entity C which is repeatedly consumed by more than one reaction of the RAF set. entities of a maxRAF set are eciently produced, then M (t) ! 0 for t ! 1 and m = 0. However, the de - s M nition of a RAF set does not ensure that all the entities A. Non interacting nets associated with such a set are present in large amount during the evolution of the system. For instance, a non- We start investigating the non interacting nets by sim- food entity that is a source for a transition of a RAF ulating the evolution of three isolated RAF sets that con- set could be continuously consumed by that transition as stitute the collection, in order to obtain the dynamics soon as it is produced, resulting in a uctuating evolution that will be the benchmark for the evolution of the com- of its number of tokens. For this reason, we consider a posite nets. FIG. 2 shows the time evolution of the num- less restrictive condition than m = 0 for the emergence ber of tokens of non-food entities for the isolated RAF set of a maxRAF set. In particular, for large t, we require represented in FIG. 1. Hereafter, we refer to this simula- a strictly positive concentration for all non-food entities tion as \reference", since it will be used as a benchmark and a high concentration for most of them. This implies 1 for the other simulations. that the relation < 1 holds for all the terms con- 1+x It is evident that, after a transient time of approxi- tributing to the calculation of M , and that the relation mately 0:06 time units, all the entities associated with 0 holds for most of them. Thus, we introduce the 1+x the RAF set grow in number, with the exception of en- following condition: tity C : C is the only entity in the set to be a source for more than one transition (FIG. 1). The green line m < 1: (4) in FIG. 3 shows the corresponding trend of M (t). As We assume that, if condition (4) holds, the entire expected, after the same transient time of  0:06 time maxRAF set emerges. units, M (t) decreases down to values close to zero, the entire maxRAF set appears and condition (4) is satis ed Note, moreover, that di erent growing rates among en- (FIG. 7). tities of a RAF set correspond to di erent e ective ring rates of the transitions. Therefore, even if a maxRAF We nd that the asymptotic dynamics of simple iso- R set appears, the time  necessary to perform all the lated RAF sets is not a ected by changing the initial transition of R can exhibit di erent slopes during the state. This is obvious from FIG. 3 and FIG. 4 by com- evolution of the net, based on the various e ective rates paring the trends of M (t) and  obtained for the same s i of subsets of R. We use the slope m of the straight line RAF set for various initial conditions. In particular, we y (i) = m i +q that approximates  for i ! 1 in order perform di erent simulations by setting x (t = 0) equal i s to compare the eciency of the (total) self-production of to a random number less than 10 for all non-food enti- di erent RAF sets. We summarise the results obtained ties of the maxRAF set, and by setting x (t = 0) = 10 from the simulated composition operations and the dif- for only those entities associated with a particular closed ferent RAF sets of our collection through a scatter plot RAF set. For all the RAF sets in the collection, the re- showing the values (m ; m ) averaged over independent sulting values of m and m are in agreement with those M  M runs (FIG. 7). corresponding to the initial conditions described by Eq. 7 FIG. 3 Total production of non-food entities by isolated nets. FIG. 4 Cycles of transitions completed by isolated nets. M (t) obtained in a simulation run of the isolated RAF set obtained for the isolated RAF set shown in FIG. 1 with di er- shown in FIG. 1 for di erent initial conditions: solid green ent initial conditions: solid green line (reference simulation): line (reference simulation): x (t = 0) = 100 if s 2 F , x (t = x (t = 0) = 100 if s 2 F , x (t = 0) = 0 otherwise; dotted ma- s s s s 0) = 0 otherwise; dotted magenta line: x (t = 0) = 100 genta line: x (t = 0) = 100 if s 2 F , x (t = 0) = rand(0; 10 ) s s s if s 2 F , x (t = 0) = rand(0; 10 ) otherwise; dashed blue otherwise; dashed blue line: x (t = 0) = 100 if s 2 F , s s 4 4 line: x (t = 0) = 100 if s 2 F , x (t = 0) = 10 if s 2 x (t = 0) = 10 if s 2 fE; H; N; Og, x (t = 0) = 0 otherwise. s s s s fE; H; N; Og, x (t = 0) = 0 otherwise. After a transient time The similar slope of  for t ! 1 associated with di erent s i of t  0:006 (arbitrary units) M (t) always takes values close tested con gurations indicates that, after a transient time, to zero, indicating that the maxRAF set has emerged for all the eciency of performing all the transitions of the RAF set the di erent tested con gurations. Dash-dotted red line: net does not depend on the initial conditions. Red circle: the net that failed to be a RAF set, obtained by switching o the that failed to be a RAF set, obtained by switching o the catalysis (D; 8) in the RAF set shown in FIG. 1. In this case, catalysis (D; 8) in the RAF set shown in FIG. 1. In this case, M (t) is always greater than 1, indicating that at least one the net is not able to perform all its transitions. non-food entity is not produced by transitions of the system. 2 cannot proceed for a certain time interval of 10 time (3) (FIG. 7). Conversely, the dynamics emerging in a steps. Hereafter, we refer to this particular con guration net that failed to be a RAF set is signi cantly di erent as the \delayed con guration". (FIG. 3 and FIG. 4, red lines). We nd that composition operations do not have any These results suggest once again that simple RAF sets impact on the emergence of the maxRAF sets for any have an e ective advantage in self-reproduction over non- choice of S that does not include food entities (FIG. 7). RAF set. Also, the structure of RAF sets alone is not However, a primordial form of biological interactions can sucient to guarantee the presence and the dynamical be established; namely, facilitation and cheating. In par- selectability of di erent long-term behaviours. Using the ticular, in nets composed through operations CO and dynamics of isolated RAF sets as reference, we can now CO , in owing of external entities can facilitate the ap- II move on to the dynamics of composed RAF sets. pearance and the sustenance of a RAF set, improving its production eciency. At the same time, withdrawing entities produced by a RAF set can counteract its pro- duction. These aspects are well highlighted by various B. Composite nets trends of the generated m due to di erent conditions. In order to compose nets, we choose ve di erent sets On the other hand, as expected, composition opera- S : the set of places belonging to the food set, the set tions involving the food set have a major role on the of places non belonging to the food set, the set of places emergence of the maxRAF set. FIG. 5 and FIG. 6 show corresponding to the molecules of length l = l + 1 and the behaviour of M (t) and  associated with two RAF f s i l  l + 1 of the BPM and the set of places that are not sets that are composed by operations CO , CO and f I II input places for spontaneous transitions (not selected for CO for S = F . Similar trends are obtained from III operation CO ). We compose copies of the RAF sets the composition of the other RAF sets in our collection. II according to the composition operations CO , CO and Results show that operations CO and CO prevent the I II I II CO . The initial states of the nets are set according entire maxRAF set in at least one of the two copies from III to Eq. (3). Moreover, for operation CO , simulations emerging. In particular, we nd that, if the rate  of III f are performed in which the transitions belonging to net transitions allowing the ux of food elements from net 1 8 FIG. 5 Total production of non-food entities by composite FIG. 6 Cycles of transitions completed by composite nets. nets. M (t) is obtained for two copies of the RAF set shown is obtained for two copies of the RAF set shown in FIG. 1 in FIG. 1 with di erent composition operations. For each with di erent composition operations. For each composite composite net, two M (t) are calculated, each from the enti- net, two  are calculated, each from the transitions associated s i ties associated with the original copies. Dash-dotted magenta with the textcolorredoriginal copies. Dash-dotted magenta 5 5 line: operation CO , net 1,  = 10  . Dashed cyan line: line: operation CO , net 1,  = 10  . Solid yellow line: I c I c f f operation CO , net 1,  = 10  . Solid yellow line: oper- operation CO , net 1. Large cyan circle: operation CO , II f c III II ation CO , net 1. Dash-dotted violet line: operation CO , net 1,  = 10  . Intermediate violet circle: operation CO , III I f c I 5 5 net 2,  = 10  . Solid blue line: operation CO , net 2, net 2,  = 10  . Intermediate blue circle: operation CO , f c II f c II 5 5 = 10  . Dotted dark-yellow line: operation CO , net net 2,  = 10  . Small dark-yellow circle: operation CO , f c II f c II 2, delayed. For all composite operations, M (t) of net 2 does net 2, delayed. For all composite operations, net 2 is not able not satisfy the condition stated by Eq. (4) that results in the to execute all its transitions. For operation CO , both net II emergence of the maxRAF set. For operation CO , both the 1 and net 2 do not perform a complete cycle of transition of II maxRAF sets of net 1 and net 2 do not emerge. the RAF set. The observed dynamics lead to some important con- to net 2 is low enough ( < 10  ), operation CO pre- f c I siderations: rst, it is clear that the actual availability of vents the emergence of the maxRAF set in net 2, while resources is a crucial element in the dynamical realisation net 1 exhibits the same dynamics of the isolated net, as of a RAF set, and the rate at which food elements enter can be concluded from FIG. 7. On the other hand, for the net is as relevant as the de nition of the food set it- the same composition operation, if the rate  is su- self, as expected. Moreover, the di erences emerging due ciently high (  10  ), the ow of food elements is f c to the impact of the CO and CO operations suggest such that the maxRAF set of net 1 does not contain avail- I II that the complexes play an important role in the dy- able resources to perform all its transitions, while net 2 namics of the RAF sets, even if they are de ned starting evolves as if it is isolated and able to drawn food directly from the single entities. We will investigate these points from the environment (FIG. 7). For operation CO , we II in a following work. Finally, as previously observed in have achieved signi cantly di erent results. In this case, (Hordijk et al., 2018b), biological interactions di erent we nd di erent threshold values  = ( ;  ) (de- f f1 f2 from competition among RAF sets are plausible. pending on the speci c RAF set of the collection) such that the emergence of the maxRAF set in net 2 is pre- It is also intriguing that an e ective competition can vent for    , while the opposite situation is ob- emerge if two nets share the same food, as in case of f f1 tained for    . Moreover, for intermediate values composition operation CO and S . In particular, we f f2 III <  <  , no maxRAF set emerges but only some observed that in the delayed con guration, the presence f1 f f2 of the closed RAF sets (FIG. 5 and FIG. 6). There- of the maxRAF of net 1 prevents the emergence of the fore, for this range of values, the ux of entities is such maxRAF of net 2. In fact, once the maxRAF of net 1 that both nets 1 and 2 have enough food elements to has had enough time to appear, the number of tokens of re transitions and perform (complementary) subsets of its associated entities increases. Since the e ective rate the maxRAF set, namely, the closed RAF sets. Further- of a transition is proportional to the number of tokens more, the stochastic nature of the ow process allows the of its sources, entities of net 1 have an higher chance of emergence of di erent closed RAF sets in each run, hence reacting with respect to their counterparts belonging to showing the possible selectability of asymptotic dynamics the delayed net 2. Most of the food elements are there- for composite nets. fore consumed by transitions of net 1. Once activated, 9 only some of the transitions of net 2 are able to be per- formed eciently, leading to the emergence of only some of the closed RAF sets which constitute the maxRAF set of net 2. Di erent runs show that the emerging closed RAF sets can vary due to the stochastic nature of the system evolution, thus guaranteeing the selectability of the di erent long-term behaviours. This result is in contrast with the previous ones where it has been observed that isolated RAF sets are not able to experiment di erent asymptotic dynamics. In fact, composing two RAF sets using operation CO produces III a composite net in which all transitions form a (larger) RAF set (see the inclusion conditions, Section III.B), and the e ect of the delay can be seen as a selection of a particular initial state. However, the same composite net is not able to produce competition if the delay is not introduced. Also, an e ective competition between FIG. 7 Impact of composition operations on the emergence closed RAF sets has not been observed in an isolated of the maxRAF set. On the x-axis: slope m of the straight net with initial conditions containing RAF sets already line tting  for t ! 1, averaged over 100 independent runs. emerged at time t = 0 (FIG. 7, FIG. 3 and FIG. 4). On the y-axis: m = M (t), averaged over 100 independent M s We suggest that key elements for this form of compe- runs. Here  is the average over time for t ! 1. For each run, both m and m are calculated for t  t , where t tition are both the structure of the composite RAF set  M f f signi es the end of the simulation run. Green triangles: RAF and the particular choice of initial conditions. In fact, in set 1 (Fig. 1). Red circles: RAF set 2. Blue squares: RAF set nets composed by operation CO and S = F , each III 3. Empty points: net 1. Filled points: net 2. Yellow shaded transition that consumes at least one food entity as a region indicates values corresponding to m ! 1, obtained source or a catalyst, always has at least one competi- for composition operations CO ; CO ; CO (delayed) and I II III tor represented by its copy. By contrast, the hierarchical S = F . structure of RAF sets does not guarantee such level of competition between di erent close RAF sets. Moreover, the delay brings the system into a state that promotes the On the other hand, composition operations involving formation of some subsets of the RAF set. The system food entities hinders the appearance of, at least, one of can hardly reach this state only through random uctua- the two original RAF sets, giving rise to possible di erent tions. The results presented in this paper show that these long-term behaviours. In particular, if the food entities two conditions allow RAF sets to have di erent accessible can be exchanged between one RAF set and another, as asymptotic dynamics. the case of composition operations CO and CO , the I II emergence of the maxRAF set of the starting nets de- pends on the rate of the exchange transitions (the ow). V. CONCLUSIONS For suciently low rates, only the RAF set capable of acquiring food directly from the environment is able to In this work we study the impact of composition opera- form. Conversely, for suciently high rates the elements tions on the dynamics of simple RAF sets. This allows us of the food are exchanged in such portions to allow only to test whether the interactions among RAF sets permit the receiving RAF set to appear. Furthermore, if com- di erent long-term dynamical behaviour of the RAF sets plexes of a reaction are involved in the exchange and not themselves, that is a necessary condition for evolvability. individual entities (as in the case of the composition op- To this aim, we generate various RAF sets starting from eration CO ), we nd an intermediate interval of ow II di erent instances of the BPM and represent RAF sets rate values within which the exchange of food elements as SPNs. Moreover, we introduce composition operations between the nets allows the emergence of just some of that, acting on SPNs, correspond to interactions among the closed RAF sets that constitute the starting maxRAF RAF sets. We nd that, if the composition operations sets. For these intermediate ow rates, therefore, the nec- do not involve the food entities of a RAF set, the dy- essary evolvability conditions are met. Finally, we nd namics of the system always reaches a state in which all that only some closed RAF sets within a maxRAF set the composed RAF sets appear. However, how fast the can appear if the maxRAF set shares the food entities RAF sets emerge and their eciency in self-reproduction with its copy that has already fully emerged. The latter depends on the exchange of entities, showing that the is a relevant result, since sharing the same food set by composition operations can give rise to interactions with two RAF sets produces a maxRAF set that is the union ecological features. of the two starting RAF sets. The dynamics observed in 10 this case shows that di erent long-term behaviours are ulating the stochastic time evolution of coupled chemical reactions. Journal of computational physics, 22(4), 403-434. possible for a single RAF set, at least as long as the sys- Gillespie, D. T., 1977. Exact stochastic simulation of cou- tem is in a particular initial state and the subsets of the pled chemical reactions. The journal of physical chemistry, RAF set compete with each other for each reaction that 81(25), 2340-2361. needs food elements. Gillespie, D. T., 2007. Stochastic simulation of chemical ki- In previous works it was theorised that separate RAF netics. Annu. Rev. Phys. Chem., 58, 35-55. networks could compete for shared resources (Hordijk Haas, P. J., 2006. Stochastic petri nets: Modelling, stability, simulation. Springer Science and Business Media. and Steel, 2014; Kau man, 2011; Vasas et al., 2012), and Higgs, P. G., Lehman, N., 2015. The RNA World: molecular a competitive dynamics was observed in spatially sepa- cooperation at the origins of life. Nature Reviews Genetics, rate RAF sets (Hordijk et al., 2018b). Results presented 16(1), 7-17. in this paper con rm the possible evolvability of a sys- Hordijk, W., Steel, M., 2004. Detecting autocatalytic, self- tem of RAF sets separated into compartments. We also sustaining sets in chemical reaction systems. Journal of the- noticed that isolated (composite) RAF sets can experi- oretical biology, 227(4), 451-461. ence di erent asymptotic dynamics. Furthermore, recent Hordijk, W., Kau man, S. A., Steel, M., 2011. Required levels of catalysis for emergence of autocatalytic sets in models of results show the presence of RAF sets in real biological chemical reaction systems. International journal of molec- systems (Sousa et al., 2015; Xavier et al., 2020), con rm- ular sciences, 12(5), 3085-3101. ing their biological interest. However, since all the simple Hordijk, W., Steel, M., 2012. Autocatalytic sets extended: (tested) isolated RAF sets experience the emergence of Dynamics, inhibition, and a generalization. Journal of Sys- the entire maxRAF set, the de nition of the RAF sets tems Chemistry, 3(1), 5. does not seem to be sucient for implying dynamics with Hordijk, W., Steel, M., Kau man, S. A., 2012. The structure multiple selectable long-term behaviours. of autocatalytic sets: Evolvability, enablement, and emer- gence. Acta biotheoretica, 60(4), 379-392. In order to further clarify this last point, we will ex- Hordijk, W., Steel, W., 2014. Conditions for evolvability of plore a larger ensemble of RAF sets in a forthcoming autocatalytic sets: A formal example and analysis. Origins paper. Moreover, in the future it might be interesting of Life and Evolution of Biospheres, 44(2), 111-124. to study the composition of RAF sets enclosed in pro- Hordijk, W., Steel, M., 2017. Chasing the tail: The emergence tocells, investigating how the coupling between internal of autocatalytic networks. Biosystems, 152, 1-10. networks and boundaries a ects the global dynamics of Hordijk, W., Steel, M., Dittrich, P., 2018. Autocatalytic sets the system. and chemical organizations: Modeling self-sustaining reac- tion networks at the origin of life. New Journal of Physics, 20(1), 015011. Hordijk, W., Naylor, J., Krasnogor, N., Fellermann, H., 2018. ACKNOWLEDGMENTS Population dynamics of autocatalytic sets in a compart- mentalized spatial world. Life, 8(3), 33. We would like to thank Marco Pedicini for his ideas, Hordijk, W., Steel, M., 2018. Autocatalytic networks at the insightful help and valuable discussions. We want to ac- basis of lifes origin and organization. Life, 8(4), 62. Jain, S., Krishna, S., 1998. Autocatalytic sets and the growth knowledge helpful suggestions from Marco Villani and of complexity in an evolutionary model. Physical Review Fatemeh Zahara Majidi, the latter also for improving the Letters, 81(25), 5684. manuscript with a precise and accurate proof-reading. Kau man, S. A., 1971. Cellular homeostasis, epigenesis and replication in randomly aggregated macromolecular sys- tems. J. Cybernetics 1, 71-96. REFERENCES Kau man, S. A., 1986. Autocatalytic sets of proteins. Journal of theoretical biology, 119(1), 1-24. Kau man, S. A., 1993. The origins of order: Self-organization Anderson, D. F., Kurtz, T. G., 2011. Continuous time Markov chain models for chemical reaction networks. In Design and and selection in evolution. Oxford University Press, USA. Kau man, S. A., 2011. Approaches to the origin of life on analysis of biomolecular circuits (pp. 3-42). Springer, New York, NY. earth. Life, 1(1), 34-48. Lohn, J. D., Colombano, S. P., Scargle, J., Stassinopoulos, D., Baez, J. C., Pollard, B. S., 2017. A compositional framework Haith, G. L., 1998. Evolving catalytic reaction sets using for reaction networks. Reviews in Mathematical Physics, 29(09), 1750028. geneticalgorithms. Proceedings of the 1998 IEEE Interna- tionalConference on Evolutionary Computation, Anchor- Benner, S. A., Kim, H. J., Yang, Z., 2012. Setting the stage: the history, chemistry, and geobiology behind RNA. Cold age, AK,USA. pp. 487492. Luisi, P. L., 2016. The emergence of life: from chemical origins Spring Harbor perspectives in biology, 4(1), a003541. Bernhardt, H. S., 2012. The RNA world hypothesis: the worst to synthetic biology. Cambridge University Press. Molloy, M., 1982. Performance analysis using stochastic Petri theory of the early evolution of life (except for all the oth- ers). Biology direct, 7(1), 23. nets. IEEE Transactions on computers, (9), 913-917. Mossel, E., Steel, M., 2005. Random biochemical networks: Feinberg, M., 1995. The existence and uniqueness of steady states for a class of chemical reaction networks. Archive for The probability of self-sustaining autocatalysis. Journal of theoretical biology, 233(3), 327-336. Rational Mechanics and Analysis, 132(4), 311-370. Gillespie, D. T., 1976. A general method for numerically sim- Nghe, P., Hordijk, W., Kau man, S. A., Walker, S. I., 11 Schmidt, F. J., Kemble, H., Yeates, J. A. M., Lehman, tocatalytic sets in E. coli metabolism. Journal of Systems N., 2015. Prebiotic network evolution: six key parameters. Chemistry, 6(1), 4. Molecular BioSystems, 11(12), 3206-3217. Szostak, J. W., 2017. The narrow road to the deep past: in Petri, C. A., Reisig, W., 2008. Petri net. Scholarpedia, 3(4), search of the chemistry of the origin of life. Angew. Chem. 6477. Int. Ed. Engl. 56, 11 037 - 11 043. Rasmussen, S., Chen, L., Deamer, D., Krakauer, D. C., Vasas, V., Szathmry, E., Santos, M., 2010. Lack of evolv- Packard, N. H., Stadler, P. F., Bedau, M. A., 2004. Tran- ability in self-sustaining autocatalytic networks constraints sitions from nonliving to living matter. Science, 303(5660), metabolism- rst scenarios for the origin of life. Proceedings 963-965. of the National Academy of Sciences, 107(4), 1470-1475. Rasmussen, S., Constantinescu, A., Svaneborg, C., 2016. Gen- Vasas, V., Fernando, C., Santos, M., Kau man, S. A., Sza- erating minimal living systems from non-living materials thmry, E., 2012. Evolution before genes. Biology direct, and increasing their evolutionary abilities. Philosophical 7(1), 1. Transactions of the Royal Society B: Biological Sciences, Vazquez, C. R., Silva, M., 2011. Stochastic continuous Petri 371(1701), 20150440. nets: An approximation of Markovian net models. IEEE Serra, R., Villani, M., 2019. Sustainable growth and synchro- Transactions on Systems, Man, and Cybernetics-Part A: nization in protocell models. Life, 9(3), 68. Systems and Humans, 42(3), 641-653. Sharov, A. A., 1991. Self-reproducing systems: structure, Xavier, J. C., Hordijk, W., Kau man, S. A., Steel, M., Mar- niche relations and evolution. BioSystems, 25(4), 237-249. tin, W. F., 2020. Autocatalytic chemical networks at the Smith, J. I., Steel, M., Hordijk, W., 2014. Autocatalytic sets origin of metabolism. Proceedings of the Royal Society B, in a partitioned biochemical network. Journal of Systems 287(1922), 20192377. Chemistry, 5(1), 2. Sousa, F. L., Hordijk, W., Steel, M., Martin, W. F., 2015. Au- http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nonlinear Sciences arXiv (Cornell University)

Impact of composition on the dynamics of autocatalytic sets

Ravoni, Alessandro
Nonlinear Sciences , Volume 2020 (2009) – Sep 13, 2020
Download PDF
11 pages

Loading next page...
 
/lp/arxiv-cornell-university/impact-of-composition-on-the-dynamics-of-autocatalytic-sets-U6vzt2pIlz
ISSN
0303-2647
eISSN
ARCH-3338
DOI
10.1016/j.biosystems.2020.104250
Publisher site
See Article on Publisher Site

Abstract

Alessandro Ravoni Department of Mathematics and Physics, University of Roma Tre, Via della Vasca Navale 84, 00146 Rome, Italy Autocatalytic sets are sets of entities that mutually catalyse each other's production through chemical reactions from a basic food source. Recently, the re exively autocat- alytic and food generated theory has introduced a formal de nition of autocatalytic sets which has provided promising results in the context of the origin of life. However, the link between the structure of autocatalytic sets and the possibility of di erent long-term behaviours is still unclear. In this work, we study how di erent interactions among autocatalytic sets a ect the emergent dynamics. To this aim, we develop a model in which interactions are presented through composition operations among networks, and the dynamics of the networks is reproduced via stochastic simulations. We nd that the dynamical emergence of the autocatalytic sets depends on the adopted composition operations. In particular, operations involving entities that are sources for autocatalytic sets can promote the formation of di erent autocatalytic subsets, opening the door to various long-term behaviours. Keywords: Autocatalytic sets; Origin of life; Network Composition; Stochastic Petri nets; Binary polymer model I. INTRODUCTION Moreover, RAF theory has successfully proved that RAF sets have a hierarchical structure, where the largest set of reactions within a CRS satisfying RAF prop- The exact sequence of events that led to the formation erty (the so-called maxRAF) has many subsets that of the rst living organisms from non-living matter is are smaller RAF sets themselves (Hordijk et al., 2012; still a topic under debate (Benner et al., 2012; Bernhardt, Hordijk and Steel, 2014, 2018). The latter is a pecu- 2012; Luisi, 2016; Rasmussen et al., 2004; Szostak, 2017). liarity that makes RAF sets potentially suitable for ex- On the other hand, some particular qualities of early periencing adaptive evolution (a feature that is gener- organisms are commonly accepted and are well known. ally referred to as evolvability); i.e., to collect evolution- Among these, we can point out the self-replication ability ary changes bene cial for survival and reproduction in a of the early life forms (Higgs and Lehman, 2015; Kau - given environment. Indeed, it has been argued that the man, 1993; Luisi, 2016; Nghe et al., 2015; Rasmussen et autocatalytic subsets present within the structure of the al., 2016). In this scenario, the autocatalytic sets (ASs) maxRAF could be the elementary units on which natu- are of great interest. Introduced by Kau man (Kau - ral selection can act (Hordijk and Steel, 2014; Hordijk et man, 1971, 1986, 1993), ASs are sets of entities capable al., 2018b; Vasas et al., 2012): the availability of sponta- of spontaneous emergence and self-reproduction through neous reactions would allow the occurrence of mutations catalytic reactions, starting from a nite set of entities and, consequently, the appearance of novel autocatalytic assumed to be available in the environment. There are subsets able to replicate themselves with di erent rates several de nitions of ASs in the literature (see, for in- and competing with each other. stance, (Jain and Krishna, 1998; Sharov, 1991)). Re- cently, Hordijk et al. (2011) have introduced the notion of However, rst results show that the asymptotic dy- re exively autocatalytic and food generated (RAF) sets, namics of simple RAF sets eventually reaches the state a formal de nition of ASs in the framework of chemi- in which all the reactions of the maxRAF occur catalyt- cal reaction systems (CRSs) (see Section II for de ni- ically (Hordijk et al., 2018b; Vasas et al., 2012). In this tions). Properties of RAF sets have been studied by var- state, all the elementary autocatalytic units coexist with- ious authors, and among the most important results (see out e ectively competing with each other, thus leaving (Hordijk and Steel, 2017, 2018) for more details) note the no room for adaptive evolution (Hordijk et al., 2018b; implementation of a polynomial time algorithm able to Vasas et al., 2010, 2012). The evolvability of RAF sets identify the presence of a RAF set in a general network can be restored by embedding them into compartments of interacting entities (Horidjk and Steel, 2004; Hordijk and allowing the sharing of resources and the exchange et al., 2011; Hordijk and Steel, 2012) and the detection of chemical molecules (Hordijk et al., 2018b; Kau man, of an autocatalytic structure in the metabolic network of 2011; Serra and Villani, 2019; Vasas et al., 2012). In fact, Escherichia Coli (Sousa et al., 2015) and ancient anaero- through numerical simulations, it has been observed that bic autotrophs (Xavier et al., 2020). RAF sets enclosed in semipermeable protocells can reach arXiv:2009.05958v1 [nlin.AO] 13 Sep 2020 2 di erent asymptotic states (Serra and Villani, 2019), and { ; is a pseudo-entity representing the environment that spatially separated RAF sets consuming the same and s 2 S is a food entity if there exists a reaction food source can give rise to di erent combinations of r 2 R such that (;; r) 2 E and (r; s) 2 E . r r competing autocatalytic subsets (Hordijk et al., 2018b), suggesting that the evolvability of RAF sets is related to Note that E is called the spontaneous reactions set. the interactions among RAF sets themselves. The edges (s; r) 2 E can be interpreted as source entity In this work we investigate this latter point. In par- s is consumed by reaction r, while in edges (r; s) 2 E ticular, we study the role of various interactions among entity s is produced by reaction r. i denotes a reaction RAF sets in order to understand how these interactions such that (;; i ); (i ; s) 2 E , implying an input reaction s s r a ect the emergent dynamics. To this aim, we use the producing a food entity. Moreover, we introduce out ow stochastic Petri nets (SPNs) formalism (Haas, 2006; Mol- reactions o such that (s; o ); (o ;;) 2 E . Thus, the s s s r loy, 1982) to represent and evolve RAF sets. Further- CRS is a ow reactor that allows in ow and out ow of more, we introduce some composition operations acting entities from and towards the environment. on nets, which correspond to di erent interactions among Let R represent a subset of R. The closure cl (F ) RAF sets. In this framework, assuming that the entire is de ned to be the (unique) minimal subset of S that maxRAF set always emerges in an isolated RAF set, contains F together with all entities that can be produced our goal is to nd some composition operations under from F by repeated applications of reactions in R. Note which the dynamical appearance of the maxRAF is not that cl (F ) is well de ned and nite (Horidjk and Steel, invariant. This means that the corresponding interac- 2004). Given a CRS (S; R; C; F ), a RAF set is a set of tion causes only some of the maxRAF subsets to emerge, reactions R  R (and associated entities) that satis es allowing the existence of multiple long-term behaviours the following properties: required for the evolvability of RAF sets. The paper is organised as follows. In Section II we in- 0 { Re exively autocatalytic: for each reaction r 2 R troduce the de nitions of RAF sets and SPNs. In Section there exists at least one entity s 2 cl (F ) such III we describe the model we use to evolve nets and we that (s; r) 2 C ; introduce the composition operations. In Section IV we present and analyse the results obtained by simulating { F -generated: for each reaction r 2 R and for each the dynamics of various composed RAF sets. Finally, in entity s 2 S such that (s; r) 2 E , it is s 2 cl (F ). r R Section V we discuss and demonstrate the conclusions. Thus, a RAF set is a set of reactions able to catalytically produce all its source entities starting from a suitable II. BACKGROUND food set. It is also possible to de ne the closure of a set A. Re exively autocatalytic and food generated sets of reactions, introducing the notion of closed RAF sets (Smith et al., 2014). Given a CRS (S; R; C; F ), a subset R of R is said to be a closed RAF set if: In RAF theory (Horidjk and Steel, 2004; Hordijk et al., 2011), a network of interacting entities is represented by a CRS. Following previous de nitions (Horidjk and { R is a RAF set; Steel, 2004; Hordijk and Steel, 2017; Lohn et al., 1998), we introduce a CRS as a tuple (S; R; C ) such that: { 8 r such that all its source entities and at least one catalyst are either part of the set F or are produced { S is a set of entities; 0 0 by a reaction from R , it is r 2 R . { R is a set of reactions,  ! , where ;  2 S are the reactants and products of a reaction, respec- Authors of (Hordijk et al., 2018a) describe a procedure tively; to detect closed RAF sets in a generic CRS. It has been argued that closed RAF sets are associated with { C is a catalysis set, that is, a set of pairs f(s; r); s 2 the attractors of the dynamics of a CRS and there- S ; r 2 Rg indicating the entity s as the catalyst of fore represent the relevant units capable of experienc- reaction r. ing adaptive evolution (Hordijk et al., 2018a,b; Smith et al., 2014; Vasas et al., 2012). FIG. 1 shows an exam- We also de ne a food set F  S such that entities s 2 F ple of a RAF set and its constituent closed RAF sets. are assumed to be available from the environment. We We exemplify our point by introducing a set of entities describe a CRS as a bipartite graph such that: S = fA; B; C; D; E; F; G; H; I; L; M; N; O; Pg and a food set F = fA; B; D; F; I; Mg. In this example, the RAF set { nodes are of two kinds V = S [ R; is composed of the following reactions (an entity above { edges are of two kinds E = E [ C ; the arrow of a reaction indicates the catalyst associated r 3 B. Stochastic Petri nets A Petri net consists of (see Petri and Reisig (Petri and Reisig, 2008) for further details): { a nite set of place P ; { a nite set of transitions W ; { functions b; e : P  W ! N. Here b(p; w) and e(p; w) are the number of edges from place p to transition w and from transition w to place p, respectively. The sets b(w)  P and e(w)  P are the sets of places connected to transition w by at least one edge. A marking X of a Petri net is a map X : P ! N that assigns a number of tokens to each place. In fact, a FIG. 1 RAF set example. Entities are represented by let- marking X identi es a state of the system in the space of ters (dark-green for food-entities, black for non-food entities). possible con gurations of tokens available in each place. The food set is surrounded by a dark-green dashed rectangle. With x , we indicate the number of tokens of place p Reactions are displayed by coloured squares. A black arrow available in marking X . Firing a transition w consumes emerging from a letter towards a square (from a square point- b(p; w) tokens from each of its input places p 2 b(w), ing at a letter) indicates the corresponding entity is a source (a product) for that reaction. Red dashed arrows indicate and produces e(p ; w) tokens in each of its output places catalysis. The maxRAF set consists of three closed RAF sets: p 2 e(w). For each marking X , a transition w is enabled (1) (2) R = fr ; r g (green squares), R = fr ; r ; r ; r g (ma- 1 3 1 3 5 8 (it may re) if there are enough tokens in its input places (2) genta and green squares), R = fr ; r ; r ; r ; r ; r g (blue 1 3 2 4 6 7 making the consumption possible. This shall occur, if and green squares). and only if X (p)  b(p; w), 8p 2 P . A stochastic Petri net (Haas, 2006; Molloy, 1982) is a Petri net for which each transition is equipped with a (possibly marking de- with that reaction): pendent) rate for the exponentially distributed transition ring times.  denotes the set of ring rates of a SPN. r : A + B ! C Note that the evolution of a SPN with exponentially dis- tributed transition rates is isomorphic to continuous-time r : A + D ! E Markov chain (Molloy, 1982). Petri net formalism provides a suitable environment r : C + F ! G for studying the composition of networks, with both a r : B + E ! H 4 computational and theoretical approach; the latter, in P particular, in the context of category theory (Baez and r : C + I ! L Pollard, 2017). Moreover, this formalism can describe r : C + M ! N nets with di erent dynamics (Vazquez and Silva, 2011). The authors will investigate these aspects in a forthcom- r : D + M ! O ing work. r : F + L ! P III. THE MODEL The maxRAF set consists of three closed RAF sets: A. Building the net (1) (2) (2) R = fr ; r g, R = fr ; r ; r ; r g, R = 1 3 1 3 5 8 fr ; r ; r ; r ; r ; r g. Note that the closed RAF set Given a CRS (S; R; C; F ) (or a set of reactions and 1 3 2 4 6 7 (1) R = fr ; r g is also a constructively autocatalytic and associated entities that is a RAF set), we build a SPN by 1 3 food generated (CAF) set (Mossel and Steel, 2005); i.e. adding a place p for each species s 2 S and a transition w a RAF set for which reactions can take place on the con- for each reaction r 2 R such that b(w) = (r) and e(w) = dition that their catalysts are already made available by (r), where (r) and (r) are the set of all sources and the occurrence of catalysed reactions (starting from F ). targets entities of edges (s; r); (r; s) 2 E , respectively. More restrictive conditions of this kind result in higher Note that with this notation we consider both in owing rates of catalysis that form a CAF set (Mossel and Steel, and out owing transitions. Moreover, for each catalysis 2005). All the RAF sets studied in this work contain one (s; r) 2 C we add a transition w such that b(w) = (r)[s RAF subset which in turn is a CAF set. and e(w) = (r) [ s. Hereafter, we use S and R to 4 indicate both species and reactions of a CRS and places work (Feinberg, 1995). The authors are currently inves- and transitions of a SPN. The rates (r) associated with tigating this aspect in an upcoming work. each transition r are marking dependent rates: With this limitation, all the spontaneous reactions of our model are binary reactions (i.e., reactions with two (r) = h (X ) : (1) r r reactants, possibly of the same entity type), while all the catalysed reactions are ternary reactions. Even if Here,  is a xed constant depending on the type of its ternary reactions are rare, they can represent a rst ap- corresponding reaction in the CRS ( = f ;  ;  ;  g r s c i o proximation of two or more elemental reactions, such as in which  speci es spontaneous, catalysed, in owing the sequence of reactions of an enzyme catalysis (Gille- and out owing reactions, respectively) and h (M ) is a spie, 2007). value proportional to the number of combinations of to- kens available in the input places of transition r in the state X . Thus, explicitly, we shall have: B. Composition b(j; r) j j h (X ) = ; (2) We model interactions between CRSs as composition jb(r)j1 b(j; r) operations between SPNs. First, note that RAF sets sat- isfy the following conditions (hereafter, we refer to these where the product is among all the input places of tran- as the inclusion conditions) (Horidjk and Steel, 2004): sition r. We set functions b and e such that the in owing transitions do not consume tokens of the pseudo-entity if R is RAF in (S ; R ; C ; F ), it is RAF also 1 1 1 1 ; and produce a xed value of tokens of the food enti- in (S ; R ; C ; F ), if conditions S  S ; R 2 2 2 2 1 2 1 ties, while the out owing transitions consume a token of R ; C  C ; F  F are satis ed; 2 1 2 1 2 the out owing entities and do not produce tokens of the 0 0 if R is RAF (S ; R ; C ; F ) and R is RAF in 1 1 1 1 1 2 pseudo-entity ;. Thus, the rate of out owing transitions 0 0 (S ; R ; C ; F ), then R [ R is RAF in (S [ 2 2 2 2 1 is proportional to the amount of tokens of the out owing 1 2 S ; R [ R ; C [ C ; F [ F ). 2 1 2 1 2 1 2 entity, while the rate of in owing transitions is indepen- dent of the state of the system. It is noteworthy that Thus, if composition does not remove entities from the the in owing of food elements still remains a stochastic food set or reactions belonging to a RAF set, it will not event. We add in owing transitions for entities not be- have any impact on its RAF property. However, the dy- longing to the originary food set F (setting the rate of namical behaviour of the composed system can be, gen- such transition equal to zero), if required for the purpose erally, di erent from that of the starting one. of composition between nets (see Section III.B). The dy- Let (S ; R ; b ; e ;  ) and (S ; R ; b ; e ;  ) be two 1 1 1 1 1 2 2 2 2 2 namics of the obtained SPN is described by the stochastic SPNs and let I; O be subsets of their in owing and out- mass action kinetics, that is the classical dynamics used owing transitions sets. Let be the equivalence relation to represent chemical reactions, assuming a well-stirred such that s  s if i 2 I and o 2 O for some choice of s s system (Anderson and Kurtz, 2011). I; O. S denotes the set of places identi ed by relation The CRSs used in this work are generated according . We de ne the following composition operations: to the binary polymer model (BPM) (Kau man, 1986). The BPM produces a CRS where the entities set S con- CO : (S ; S ) ! S = S [ S ; I 1 2  1 2 sists of all bit strings up to (and including) a maximum (R ; R ) ! R = (R [ R [ R )n (O [ I ); 1 2  1 2 I 1 2 length N , and the reaction set R consists of condensation (b ; b ) ! b = b [ b [ b ; 1 2  1 2 I and cleavage reactions. Condensation reaction is a con- (e ; e ) ! e = e [ e [ e ; catenation of two bit strings resulting in a longer string, 1 2  1 2 I and cleavage reaction cats a bit string into two smaller ( ;  ) !  =  [  [  ; 1 2  1 2 I ones. The food set is represented by all entities with 0 0 R := frj b(r) = s; e(r) = s ; 8 s 2 S ; s 2 I 1 a length less than or equal to a xed length l (we set S such that s  s g; l = 2), and each entity can be a catalyst of each reaction := f(r)j (r) = h (X ) ; 8 r 2 R g. with a certain probability xed a priori. I r f I We chose to only allow the condensation reactions to CO : (S ; S ) ! S = S [ S ; II 1 2  1 2 occur in the system. Note that the technique we use is (R ; R ) ! R = (R [ R [ R )n (O [ I ); 1 2  1 2 II 1 2 also applicable in the case where cleavage reactions are allowed. In fact, none of the operations we introduce (b ; b ) ! b = b [ b [ b ; 1 2  1 2 II in Section III.B is a ected by the reversibility of the (e ; e ) ! e = e [ e [ e ; 1 2  1 2 II network reactions. Including cleavage reactions would ( ;  ) !  =  [  [  ; 1 2  1 2 II produce networks, in principle, with di erent dynamics, 0 00 00 since a network with irreversible reactions can have dif- R := frj b(r) = b(r ); e(r) = b(r ); 8 r 2 II 00 0 00 ferent topology with respect the equivalent reversible net- R such that b(r )  S and b(r )  b(r )g; 2  5 := f(r)j (r) = h (X ) ; 8 r 2 R g. It is worth to underline that, if there exists a s 2 S II r f II such that s 2 F , operations CO and CO can actu- I II CO : (S ; S ) ! S = fS t S g= ; ally modify the RAF property of net 2. In particu- III 1 2  1 2 lar, if transitions r 2 R are assumed to be spon- (I;II) (R ; R ) ! R = R [ R ; 1 2  1 2 taneous transitions ( <  ), net 2 could not be cat- f c (b ; b ) ! b = b [ b ; 1 2  1 2 alytically produced starting from the food set F . In this case, the composed net contains a RAF set R such (e ; e ) ! e = e [ e ; 1 2  1 2 0 0 that R  R  (R [ R ), with R = R if F  S . In- 1 1 2 1 ( ;  ) !  =  [  . 1 2  1 2 stead, if transitions r 2 R are assumed to be (auto) (I;II) catalysed transitions (   ), the whole composed net f c Here  is a constant value and (S ; R ; b ; e ;  ) is the shall be a RAF set. composed SPN. To summarise, all the composition operations we de- ne relate a set of places S that are input for out owing IV. RESULTS AND DISCUSSION transitions of a SPN, together with a set of places S that are input for in owing transitions of another SPN. The In this Section, we present the results regarding the formal addition of in owing transitions for places not be- impact of composition on the dynamics of RAF sets as longing to the food set enlarges the possible composition follows: we rst introduce the characteristics of simula- operations between SPNs. tions and the quantities taken into account. In Section Given a set S , operation CO adds a transition from I IV.A we present results for the non interacting nets, in S to S for each pair of places in S , while operation order to have a reference model for the interacting cases, CO adds a transition from S to S for each combina- II presented in Section IV.B. tion of places that appears as input of a transition in Starting from di erent instances of the BPM with the in owing net. Each combination corresponds to the N < 8, we use the RAF algorithm introduced in (Hordijk de nition of complex in the framework of chemical re- et al., 2011) in order to detect and select three di erent action networks (Feinberg, 1995). In fact, given a set of RAF sets, each of which contains more than one closed chemical species, a complex is de ned as a member of RAF set. Note that, even if di erent RAF sets have the the vector space generated by the species that provides same species set S, the set of reactions E will be di erent; the inputs (or the outputs) of a reaction (Anderson and i.e., various RAF sets have di erent chemistry. The RAF Kurtz, 2011; Feinberg, 1995). sets identi ed through this procedure constitute the col- Both operation CO and CO introduce a ux of en- I II lection on which we will carry out the study. Even if such tities from one net to another. The composite network a small collection cannot be taken as a solid statistical can therefore be seen as the union of two separate net- basis, it is still sucient for providing us with interest- works that evolve in parallel, communicating only with ing information. We duplicate each RAF set and let it (asymmetrical) exchange of chemical species. This could interacts with its copy by switching to representation of be, for instance, the case of two spatially separated pro- RAF set as an SPN and using one of the composition tocells, one of which can release molecules towards the operations introduced above. We simulated the dynam- other. According to this interpretation, the owing rate ics of the system using the standard Gillespie algorithm is a parameter that encompasses the characteristics (Gillespie, 1976, 1977), setting the volume of the system of the ow process (for example, cell permeability). Op- at V = 1 (arbitrary units). For each simulation, we per- eration CO , instead, merges each pair of places in S , form 100 independent runs of 10 time steps. One of III allowing transitions of the two original SPNs to operate the necessary conditions for a RAF set is the ability to on the glued set of places. Composing nets via opera- produce itself starting from the elements of the food set. tion CO actually produces a new single network. In Indeed, the initial state of the SPN is set such that: III this case, one can think of composite net as the result of mutations that enlarge a network (for instance, net x (t = 0) = x if s 2 F ; s 0 (3) 1), introducing new possible reactions and, consequently, x (t = 0) = 0 otherwise: new chemical species (corresponding to net 2). Here x is an arbitrary constant. The values of x ,  , 0 0 c and  are set such that the number of tokens of food i o places at t ! 1 is equal to 10 , for an SPN with in ow, Note that complexes play a major role in the framework of chem- out ow and all (and only) binary transitions having, as ical reaction networks theory. For instance, the de ciency theo- input, food places only (and ring rate  ). The value rems (Feinberg, 1995) are able to predict whether the dynamics of  is xed at  =  =10, while  varies such that of a large class of networks will have a stationary distribution, s s c f 1 6 starting from the topology of the reaction graph having com-  2 [ 10 ;  10 ]. It is noteworthy that the values f c c plexes as nodes. of these parameters are not taken from \in vivo" data, 6 but they have phenomenological motivations. Thus, al- though we can reasonably generalise the characteristics of the dynamics, quantities such as the species' produc- tion rate or the time evolution of the concentrations may di er from those in other similar stochastic simulations (Hordijk and Steel, 2012; Hordijk et al., 2018b). We focus our attention on the e ective appearance of a maxRAF set during the evolution of the system. In particular, we introduce the following quantities: 1) M (t) = , 8s 2= F ; 1+x (t) 2)  = minftj n(r)  i, 8r 2 Rg. Here n(r) is the number of executions of transition r, and R is the maxRAF set. Note that the natural con- dition x (t)  0 implies that each term contributing to FIG. 2 Isolated RAF set. Number of tokens over time of the computation of M (t) can possess, at most, the value non-food entities obtained in a simulation run of the isolated one. Both M (t) and  are calculated for each single s i RAF set shown in FIG. 1. After a transient time of t net that forms the composed net. Let m = M (t) be M s 0:006 (arbitrary units) all the entities are present in a large the mean value of M (t) for t ! 1. If all the non-food amount, with the exception of the entity C which is repeatedly consumed by more than one reaction of the RAF set. entities of a maxRAF set are eciently produced, then M (t) ! 0 for t ! 1 and m = 0. However, the de - s M nition of a RAF set does not ensure that all the entities A. Non interacting nets associated with such a set are present in large amount during the evolution of the system. For instance, a non- We start investigating the non interacting nets by sim- food entity that is a source for a transition of a RAF ulating the evolution of three isolated RAF sets that con- set could be continuously consumed by that transition as stitute the collection, in order to obtain the dynamics soon as it is produced, resulting in a uctuating evolution that will be the benchmark for the evolution of the com- of its number of tokens. For this reason, we consider a posite nets. FIG. 2 shows the time evolution of the num- less restrictive condition than m = 0 for the emergence ber of tokens of non-food entities for the isolated RAF set of a maxRAF set. In particular, for large t, we require represented in FIG. 1. Hereafter, we refer to this simula- a strictly positive concentration for all non-food entities tion as \reference", since it will be used as a benchmark and a high concentration for most of them. This implies 1 for the other simulations. that the relation < 1 holds for all the terms con- 1+x It is evident that, after a transient time of approxi- tributing to the calculation of M , and that the relation mately 0:06 time units, all the entities associated with 0 holds for most of them. Thus, we introduce the 1+x the RAF set grow in number, with the exception of en- following condition: tity C : C is the only entity in the set to be a source for more than one transition (FIG. 1). The green line m < 1: (4) in FIG. 3 shows the corresponding trend of M (t). As We assume that, if condition (4) holds, the entire expected, after the same transient time of  0:06 time maxRAF set emerges. units, M (t) decreases down to values close to zero, the entire maxRAF set appears and condition (4) is satis ed Note, moreover, that di erent growing rates among en- (FIG. 7). tities of a RAF set correspond to di erent e ective ring rates of the transitions. Therefore, even if a maxRAF We nd that the asymptotic dynamics of simple iso- R set appears, the time  necessary to perform all the lated RAF sets is not a ected by changing the initial transition of R can exhibit di erent slopes during the state. This is obvious from FIG. 3 and FIG. 4 by com- evolution of the net, based on the various e ective rates paring the trends of M (t) and  obtained for the same s i of subsets of R. We use the slope m of the straight line RAF set for various initial conditions. In particular, we y (i) = m i +q that approximates  for i ! 1 in order perform di erent simulations by setting x (t = 0) equal i s to compare the eciency of the (total) self-production of to a random number less than 10 for all non-food enti- di erent RAF sets. We summarise the results obtained ties of the maxRAF set, and by setting x (t = 0) = 10 from the simulated composition operations and the dif- for only those entities associated with a particular closed ferent RAF sets of our collection through a scatter plot RAF set. For all the RAF sets in the collection, the re- showing the values (m ; m ) averaged over independent sulting values of m and m are in agreement with those M  M runs (FIG. 7). corresponding to the initial conditions described by Eq. 7 FIG. 3 Total production of non-food entities by isolated nets. FIG. 4 Cycles of transitions completed by isolated nets. M (t) obtained in a simulation run of the isolated RAF set obtained for the isolated RAF set shown in FIG. 1 with di er- shown in FIG. 1 for di erent initial conditions: solid green ent initial conditions: solid green line (reference simulation): line (reference simulation): x (t = 0) = 100 if s 2 F , x (t = x (t = 0) = 100 if s 2 F , x (t = 0) = 0 otherwise; dotted ma- s s s s 0) = 0 otherwise; dotted magenta line: x (t = 0) = 100 genta line: x (t = 0) = 100 if s 2 F , x (t = 0) = rand(0; 10 ) s s s if s 2 F , x (t = 0) = rand(0; 10 ) otherwise; dashed blue otherwise; dashed blue line: x (t = 0) = 100 if s 2 F , s s 4 4 line: x (t = 0) = 100 if s 2 F , x (t = 0) = 10 if s 2 x (t = 0) = 10 if s 2 fE; H; N; Og, x (t = 0) = 0 otherwise. s s s s fE; H; N; Og, x (t = 0) = 0 otherwise. After a transient time The similar slope of  for t ! 1 associated with di erent s i of t  0:006 (arbitrary units) M (t) always takes values close tested con gurations indicates that, after a transient time, to zero, indicating that the maxRAF set has emerged for all the eciency of performing all the transitions of the RAF set the di erent tested con gurations. Dash-dotted red line: net does not depend on the initial conditions. Red circle: the net that failed to be a RAF set, obtained by switching o the that failed to be a RAF set, obtained by switching o the catalysis (D; 8) in the RAF set shown in FIG. 1. In this case, catalysis (D; 8) in the RAF set shown in FIG. 1. In this case, M (t) is always greater than 1, indicating that at least one the net is not able to perform all its transitions. non-food entity is not produced by transitions of the system. 2 cannot proceed for a certain time interval of 10 time (3) (FIG. 7). Conversely, the dynamics emerging in a steps. Hereafter, we refer to this particular con guration net that failed to be a RAF set is signi cantly di erent as the \delayed con guration". (FIG. 3 and FIG. 4, red lines). We nd that composition operations do not have any These results suggest once again that simple RAF sets impact on the emergence of the maxRAF sets for any have an e ective advantage in self-reproduction over non- choice of S that does not include food entities (FIG. 7). RAF set. Also, the structure of RAF sets alone is not However, a primordial form of biological interactions can sucient to guarantee the presence and the dynamical be established; namely, facilitation and cheating. In par- selectability of di erent long-term behaviours. Using the ticular, in nets composed through operations CO and dynamics of isolated RAF sets as reference, we can now CO , in owing of external entities can facilitate the ap- II move on to the dynamics of composed RAF sets. pearance and the sustenance of a RAF set, improving its production eciency. At the same time, withdrawing entities produced by a RAF set can counteract its pro- duction. These aspects are well highlighted by various B. Composite nets trends of the generated m due to di erent conditions. In order to compose nets, we choose ve di erent sets On the other hand, as expected, composition opera- S : the set of places belonging to the food set, the set tions involving the food set have a major role on the of places non belonging to the food set, the set of places emergence of the maxRAF set. FIG. 5 and FIG. 6 show corresponding to the molecules of length l = l + 1 and the behaviour of M (t) and  associated with two RAF f s i l  l + 1 of the BPM and the set of places that are not sets that are composed by operations CO , CO and f I II input places for spontaneous transitions (not selected for CO for S = F . Similar trends are obtained from III operation CO ). We compose copies of the RAF sets the composition of the other RAF sets in our collection. II according to the composition operations CO , CO and Results show that operations CO and CO prevent the I II I II CO . The initial states of the nets are set according entire maxRAF set in at least one of the two copies from III to Eq. (3). Moreover, for operation CO , simulations emerging. In particular, we nd that, if the rate  of III f are performed in which the transitions belonging to net transitions allowing the ux of food elements from net 1 8 FIG. 5 Total production of non-food entities by composite FIG. 6 Cycles of transitions completed by composite nets. nets. M (t) is obtained for two copies of the RAF set shown is obtained for two copies of the RAF set shown in FIG. 1 in FIG. 1 with di erent composition operations. For each with di erent composition operations. For each composite composite net, two M (t) are calculated, each from the enti- net, two  are calculated, each from the transitions associated s i ties associated with the original copies. Dash-dotted magenta with the textcolorredoriginal copies. Dash-dotted magenta 5 5 line: operation CO , net 1,  = 10  . Dashed cyan line: line: operation CO , net 1,  = 10  . Solid yellow line: I c I c f f operation CO , net 1,  = 10  . Solid yellow line: oper- operation CO , net 1. Large cyan circle: operation CO , II f c III II ation CO , net 1. Dash-dotted violet line: operation CO , net 1,  = 10  . Intermediate violet circle: operation CO , III I f c I 5 5 net 2,  = 10  . Solid blue line: operation CO , net 2, net 2,  = 10  . Intermediate blue circle: operation CO , f c II f c II 5 5 = 10  . Dotted dark-yellow line: operation CO , net net 2,  = 10  . Small dark-yellow circle: operation CO , f c II f c II 2, delayed. For all composite operations, M (t) of net 2 does net 2, delayed. For all composite operations, net 2 is not able not satisfy the condition stated by Eq. (4) that results in the to execute all its transitions. For operation CO , both net II emergence of the maxRAF set. For operation CO , both the 1 and net 2 do not perform a complete cycle of transition of II maxRAF sets of net 1 and net 2 do not emerge. the RAF set. The observed dynamics lead to some important con- to net 2 is low enough ( < 10  ), operation CO pre- f c I siderations: rst, it is clear that the actual availability of vents the emergence of the maxRAF set in net 2, while resources is a crucial element in the dynamical realisation net 1 exhibits the same dynamics of the isolated net, as of a RAF set, and the rate at which food elements enter can be concluded from FIG. 7. On the other hand, for the net is as relevant as the de nition of the food set it- the same composition operation, if the rate  is su- self, as expected. Moreover, the di erences emerging due ciently high (  10  ), the ow of food elements is f c to the impact of the CO and CO operations suggest such that the maxRAF set of net 1 does not contain avail- I II that the complexes play an important role in the dy- able resources to perform all its transitions, while net 2 namics of the RAF sets, even if they are de ned starting evolves as if it is isolated and able to drawn food directly from the single entities. We will investigate these points from the environment (FIG. 7). For operation CO , we II in a following work. Finally, as previously observed in have achieved signi cantly di erent results. In this case, (Hordijk et al., 2018b), biological interactions di erent we nd di erent threshold values  = ( ;  ) (de- f f1 f2 from competition among RAF sets are plausible. pending on the speci c RAF set of the collection) such that the emergence of the maxRAF set in net 2 is pre- It is also intriguing that an e ective competition can vent for    , while the opposite situation is ob- emerge if two nets share the same food, as in case of f f1 tained for    . Moreover, for intermediate values composition operation CO and S . In particular, we f f2 III <  <  , no maxRAF set emerges but only some observed that in the delayed con guration, the presence f1 f f2 of the closed RAF sets (FIG. 5 and FIG. 6). There- of the maxRAF of net 1 prevents the emergence of the fore, for this range of values, the ux of entities is such maxRAF of net 2. In fact, once the maxRAF of net 1 that both nets 1 and 2 have enough food elements to has had enough time to appear, the number of tokens of re transitions and perform (complementary) subsets of its associated entities increases. Since the e ective rate the maxRAF set, namely, the closed RAF sets. Further- of a transition is proportional to the number of tokens more, the stochastic nature of the ow process allows the of its sources, entities of net 1 have an higher chance of emergence of di erent closed RAF sets in each run, hence reacting with respect to their counterparts belonging to showing the possible selectability of asymptotic dynamics the delayed net 2. Most of the food elements are there- for composite nets. fore consumed by transitions of net 1. Once activated, 9 only some of the transitions of net 2 are able to be per- formed eciently, leading to the emergence of only some of the closed RAF sets which constitute the maxRAF set of net 2. Di erent runs show that the emerging closed RAF sets can vary due to the stochastic nature of the system evolution, thus guaranteeing the selectability of the di erent long-term behaviours. This result is in contrast with the previous ones where it has been observed that isolated RAF sets are not able to experiment di erent asymptotic dynamics. In fact, composing two RAF sets using operation CO produces III a composite net in which all transitions form a (larger) RAF set (see the inclusion conditions, Section III.B), and the e ect of the delay can be seen as a selection of a particular initial state. However, the same composite net is not able to produce competition if the delay is not introduced. Also, an e ective competition between FIG. 7 Impact of composition operations on the emergence closed RAF sets has not been observed in an isolated of the maxRAF set. On the x-axis: slope m of the straight net with initial conditions containing RAF sets already line tting  for t ! 1, averaged over 100 independent runs. emerged at time t = 0 (FIG. 7, FIG. 3 and FIG. 4). On the y-axis: m = M (t), averaged over 100 independent M s We suggest that key elements for this form of compe- runs. Here  is the average over time for t ! 1. For each run, both m and m are calculated for t  t , where t tition are both the structure of the composite RAF set  M f f signi es the end of the simulation run. Green triangles: RAF and the particular choice of initial conditions. In fact, in set 1 (Fig. 1). Red circles: RAF set 2. Blue squares: RAF set nets composed by operation CO and S = F , each III 3. Empty points: net 1. Filled points: net 2. Yellow shaded transition that consumes at least one food entity as a region indicates values corresponding to m ! 1, obtained source or a catalyst, always has at least one competi- for composition operations CO ; CO ; CO (delayed) and I II III tor represented by its copy. By contrast, the hierarchical S = F . structure of RAF sets does not guarantee such level of competition between di erent close RAF sets. Moreover, the delay brings the system into a state that promotes the On the other hand, composition operations involving formation of some subsets of the RAF set. The system food entities hinders the appearance of, at least, one of can hardly reach this state only through random uctua- the two original RAF sets, giving rise to possible di erent tions. The results presented in this paper show that these long-term behaviours. In particular, if the food entities two conditions allow RAF sets to have di erent accessible can be exchanged between one RAF set and another, as asymptotic dynamics. the case of composition operations CO and CO , the I II emergence of the maxRAF set of the starting nets de- pends on the rate of the exchange transitions (the ow). V. CONCLUSIONS For suciently low rates, only the RAF set capable of acquiring food directly from the environment is able to In this work we study the impact of composition opera- form. Conversely, for suciently high rates the elements tions on the dynamics of simple RAF sets. This allows us of the food are exchanged in such portions to allow only to test whether the interactions among RAF sets permit the receiving RAF set to appear. Furthermore, if com- di erent long-term dynamical behaviour of the RAF sets plexes of a reaction are involved in the exchange and not themselves, that is a necessary condition for evolvability. individual entities (as in the case of the composition op- To this aim, we generate various RAF sets starting from eration CO ), we nd an intermediate interval of ow II di erent instances of the BPM and represent RAF sets rate values within which the exchange of food elements as SPNs. Moreover, we introduce composition operations between the nets allows the emergence of just some of that, acting on SPNs, correspond to interactions among the closed RAF sets that constitute the starting maxRAF RAF sets. We nd that, if the composition operations sets. For these intermediate ow rates, therefore, the nec- do not involve the food entities of a RAF set, the dy- essary evolvability conditions are met. Finally, we nd namics of the system always reaches a state in which all that only some closed RAF sets within a maxRAF set the composed RAF sets appear. However, how fast the can appear if the maxRAF set shares the food entities RAF sets emerge and their eciency in self-reproduction with its copy that has already fully emerged. The latter depends on the exchange of entities, showing that the is a relevant result, since sharing the same food set by composition operations can give rise to interactions with two RAF sets produces a maxRAF set that is the union ecological features. of the two starting RAF sets. The dynamics observed in 10 this case shows that di erent long-term behaviours are ulating the stochastic time evolution of coupled chemical reactions. Journal of computational physics, 22(4), 403-434. possible for a single RAF set, at least as long as the sys- Gillespie, D. T., 1977. Exact stochastic simulation of cou- tem is in a particular initial state and the subsets of the pled chemical reactions. The journal of physical chemistry, RAF set compete with each other for each reaction that 81(25), 2340-2361. needs food elements. Gillespie, D. T., 2007. Stochastic simulation of chemical ki- In previous works it was theorised that separate RAF netics. Annu. Rev. Phys. Chem., 58, 35-55. networks could compete for shared resources (Hordijk Haas, P. J., 2006. Stochastic petri nets: Modelling, stability, simulation. Springer Science and Business Media. and Steel, 2014; Kau man, 2011; Vasas et al., 2012), and Higgs, P. G., Lehman, N., 2015. The RNA World: molecular a competitive dynamics was observed in spatially sepa- cooperation at the origins of life. Nature Reviews Genetics, rate RAF sets (Hordijk et al., 2018b). Results presented 16(1), 7-17. in this paper con rm the possible evolvability of a sys- Hordijk, W., Steel, M., 2004. Detecting autocatalytic, self- tem of RAF sets separated into compartments. We also sustaining sets in chemical reaction systems. Journal of the- noticed that isolated (composite) RAF sets can experi- oretical biology, 227(4), 451-461. ence di erent asymptotic dynamics. Furthermore, recent Hordijk, W., Kau man, S. A., Steel, M., 2011. Required levels of catalysis for emergence of autocatalytic sets in models of results show the presence of RAF sets in real biological chemical reaction systems. International journal of molec- systems (Sousa et al., 2015; Xavier et al., 2020), con rm- ular sciences, 12(5), 3085-3101. ing their biological interest. However, since all the simple Hordijk, W., Steel, M., 2012. Autocatalytic sets extended: (tested) isolated RAF sets experience the emergence of Dynamics, inhibition, and a generalization. Journal of Sys- the entire maxRAF set, the de nition of the RAF sets tems Chemistry, 3(1), 5. does not seem to be sucient for implying dynamics with Hordijk, W., Steel, M., Kau man, S. A., 2012. The structure multiple selectable long-term behaviours. of autocatalytic sets: Evolvability, enablement, and emer- gence. Acta biotheoretica, 60(4), 379-392. In order to further clarify this last point, we will ex- Hordijk, W., Steel, W., 2014. Conditions for evolvability of plore a larger ensemble of RAF sets in a forthcoming autocatalytic sets: A formal example and analysis. Origins paper. Moreover, in the future it might be interesting of Life and Evolution of Biospheres, 44(2), 111-124. to study the composition of RAF sets enclosed in pro- Hordijk, W., Steel, M., 2017. Chasing the tail: The emergence tocells, investigating how the coupling between internal of autocatalytic networks. Biosystems, 152, 1-10. networks and boundaries a ects the global dynamics of Hordijk, W., Steel, M., Dittrich, P., 2018. Autocatalytic sets the system. and chemical organizations: Modeling self-sustaining reac- tion networks at the origin of life. New Journal of Physics, 20(1), 015011. Hordijk, W., Naylor, J., Krasnogor, N., Fellermann, H., 2018. ACKNOWLEDGMENTS Population dynamics of autocatalytic sets in a compart- mentalized spatial world. Life, 8(3), 33. We would like to thank Marco Pedicini for his ideas, Hordijk, W., Steel, M., 2018. Autocatalytic networks at the insightful help and valuable discussions. We want to ac- basis of lifes origin and organization. Life, 8(4), 62. Jain, S., Krishna, S., 1998. Autocatalytic sets and the growth knowledge helpful suggestions from Marco Villani and of complexity in an evolutionary model. Physical Review Fatemeh Zahara Majidi, the latter also for improving the Letters, 81(25), 5684. manuscript with a precise and accurate proof-reading. Kau man, S. A., 1971. Cellular homeostasis, epigenesis and replication in randomly aggregated macromolecular sys- tems. J. Cybernetics 1, 71-96. REFERENCES Kau man, S. A., 1986. Autocatalytic sets of proteins. Journal of theoretical biology, 119(1), 1-24. Kau man, S. A., 1993. The origins of order: Self-organization Anderson, D. F., Kurtz, T. G., 2011. Continuous time Markov chain models for chemical reaction networks. In Design and and selection in evolution. Oxford University Press, USA. Kau man, S. A., 2011. Approaches to the origin of life on analysis of biomolecular circuits (pp. 3-42). Springer, New York, NY. earth. Life, 1(1), 34-48. Lohn, J. D., Colombano, S. P., Scargle, J., Stassinopoulos, D., Baez, J. C., Pollard, B. S., 2017. A compositional framework Haith, G. L., 1998. Evolving catalytic reaction sets using for reaction networks. Reviews in Mathematical Physics, 29(09), 1750028. geneticalgorithms. Proceedings of the 1998 IEEE Interna- tionalConference on Evolutionary Computation, Anchor- Benner, S. A., Kim, H. J., Yang, Z., 2012. Setting the stage: the history, chemistry, and geobiology behind RNA. Cold age, AK,USA. pp. 487492. Luisi, P. L., 2016. The emergence of life: from chemical origins Spring Harbor perspectives in biology, 4(1), a003541. Bernhardt, H. S., 2012. The RNA world hypothesis: the worst to synthetic biology. Cambridge University Press. Molloy, M., 1982. Performance analysis using stochastic Petri theory of the early evolution of life (except for all the oth- ers). Biology direct, 7(1), 23. nets. IEEE Transactions on computers, (9), 913-917. Mossel, E., Steel, M., 2005. Random biochemical networks: Feinberg, M., 1995. The existence and uniqueness of steady states for a class of chemical reaction networks. Archive for The probability of self-sustaining autocatalysis. Journal of theoretical biology, 233(3), 327-336. Rational Mechanics and Analysis, 132(4), 311-370. Gillespie, D. T., 1976. A general method for numerically sim- Nghe, P., Hordijk, W., Kau man, S. A., Walker, S. I., 11 Schmidt, F. J., Kemble, H., Yeates, J. A. M., Lehman, tocatalytic sets in E. coli metabolism. Journal of Systems N., 2015. Prebiotic network evolution: six key parameters. Chemistry, 6(1), 4. Molecular BioSystems, 11(12), 3206-3217. Szostak, J. W., 2017. The narrow road to the deep past: in Petri, C. A., Reisig, W., 2008. Petri net. Scholarpedia, 3(4), search of the chemistry of the origin of life. Angew. Chem. 6477. Int. Ed. Engl. 56, 11 037 - 11 043. Rasmussen, S., Chen, L., Deamer, D., Krakauer, D. C., Vasas, V., Szathmry, E., Santos, M., 2010. Lack of evolv- Packard, N. H., Stadler, P. F., Bedau, M. A., 2004. Tran- ability in self-sustaining autocatalytic networks constraints sitions from nonliving to living matter. Science, 303(5660), metabolism- rst scenarios for the origin of life. Proceedings 963-965. of the National Academy of Sciences, 107(4), 1470-1475. Rasmussen, S., Constantinescu, A., Svaneborg, C., 2016. Gen- Vasas, V., Fernando, C., Santos, M., Kau man, S. A., Sza- erating minimal living systems from non-living materials thmry, E., 2012. Evolution before genes. Biology direct, and increasing their evolutionary abilities. Philosophical 7(1), 1. Transactions of the Royal Society B: Biological Sciences, Vazquez, C. R., Silva, M., 2011. Stochastic continuous Petri 371(1701), 20150440. nets: An approximation of Markovian net models. IEEE Serra, R., Villani, M., 2019. Sustainable growth and synchro- Transactions on Systems, Man, and Cybernetics-Part A: nization in protocell models. Life, 9(3), 68. Systems and Humans, 42(3), 641-653. Sharov, A. A., 1991. Self-reproducing systems: structure, Xavier, J. C., Hordijk, W., Kau man, S. A., Steel, M., Mar- niche relations and evolution. BioSystems, 25(4), 237-249. tin, W. F., 2020. Autocatalytic chemical networks at the Smith, J. I., Steel, M., Hordijk, W., 2014. Autocatalytic sets origin of metabolism. Proceedings of the Royal Society B, in a partitioned biochemical network. Journal of Systems 287(1922), 20192377. Chemistry, 5(1), 2. Sousa, F. L., Hordijk, W., Steel, M., Martin, W. F., 2015. Au-

Journal

Nonlinear Sciences – arXiv (Cornell University)

Published: Sep 13, 2020

References

  • Continuous time Markov chain models for chemical reaction networks
    Anderson
  • A compositional framework for reaction networks
    Baez
  • Setting the stage: the history, chemistry, and geobiology behind RNA
    Benner
  • The RNA world hypothesis: the worst theory of the early evolution of life (except for all the others)
    Bernhardt
  • The existence and uniqueness of steady states for a class of chemical reaction networks
    Feinberg
  • A general method for numerically simulating the stochastic time evolution of coupled chemical reactions
    Gillespie
  • Exact stochastic simulation of coupled chemical reactions
    Gillespie
  • Stochastic simulation of chemical kinetics
    Gillespie