Skip to main content
PLOS ONE logoLink to PLOS ONE
. 2021 Feb 24;16(2):e0247200. doi: 10.1371/journal.pone.0247200

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: A computational perspective

Safar Vafadar 1,#, Maryam Shahdoust 2,#, Ata Kalirad 2,#, Pooya Zakeri 3,4,#, Mehdi Sadeghi 2,5,*
Editor: Bashar Ibrahim6
PMCID: PMC7904198  PMID: 33626106

Abstract

Inspired by the competition exclusion principle, this work aims at providing a computational framework to explore the theoretical feasibility of viral co-infection as a possible strategy to reduce the spread of a fatal strain in a population. We propose a stochastic-based model—called Co-Wish—to understand how competition between two viruses over a shared niche can affect the spread of each virus in infected tissue. To demonstrate the co-infection of two viruses, we first simulate the characteristics of two virus growth processes separately. Then, we examine their interactions until one can dominate the other. We use Co-Wish to explore how the model varies as the parameters of each virus growth process change when two viruses infect the host simultaneously. We will also investigate the effect of the delayed initiation of each infection. Moreover, Co-Wish not only examines the co-infection at the cell level but also includes the innate immune response during viral infection. The results highlight that the waiting times in the five stages of the viral infection of a cell in the model—namely attachment, penetration, eclipse, replication, and release—play an essential role in the competition between the two viruses. While it could prove challenging to fully understand the therapeutic potentials of viral co-infection, we discuss that our theoretical framework hints at an intriguing research direction in applying co-infection dynamics in controlling any viral outbreak’s speed.

Introduction

The “competitive exclusion principle” (CEP), also—perhaps erroneously—known as “Gause’s law” [1], is the consequence of natural selection operating on non-interbreeding populations that occupy the same ecological niche. As Darwin put it,“the competition will generally be most severe […] between the forms which are most like each other in all respects” [2] (p.320). In the simplest reading, this principle implies that, in the competition between two sympatric non-interbreeding populations over the same ecological niche, one will displace the other.

The crux of this work is to couple the CEP with the life cycle of viruses. To model how two types of viruses, B and M, co-infect a tissue, we have to make a few assumptions:

  1. Different viruses can have eclipse phases of varying lengths. The period between the initial infection and the first detectable viremia is known as the eclipse period—from the moment a virus enters the cell until it starts assembling its progenies and subsequences burst out of the cell. The duration of this period varies among different virus strains and affects the kinetic of infection (e.g., the eclipse period of SHIV lasts around a day [3] while it lasts between 7 to 8 hours in SARS-CoV [4]).

  2. Different viruses have different burst sizes, i.e., different per-cell virion particles. This assumption is reasonable, but even for SARS-CoV-2, the estimated burst size of 103 is simply based on MHV-2 data (e.g., [5]). The variation in particle-to-PFU among animal viruses [6] can be used to deduce the veracity of this assumption, but more direct data on SARS-CoV-2 is needed.

  3. The CEP applies when two virus strains compete to infect a cell. In a spatially heterogeneous environment, different populations tend to partition the environment into non-overlapping micro-environments; for one of the most famous experimental demonstrations of such a situation see [7]. However, the displacement of one of the competitors by another is inescapable when the competitors cannot adapt or construct new niches in the environment in a reasonable timescale. The notion of CEP is undisputed when it comes to animals and bacteria trying to occupy the same ecological niche, but its application to viral confection is not as unequivocal as one would imagine.

In addition, different viruses can have different incubation times. For example, the median incubation period—i.e., the period between the onset of infection and the appearance of symptoms—for SARS-CoV-2 is estimated to be 5.1 days [8] (although such estimates should be taken with a grain of salt, e.g., [9]), whereas the incubation period for the common cold is around 1–3 days, and it could even be as long a few months to few years (e.g., Rabies and AIDS) [6]. Here, we do not consider any designation to simulate the different incubation times of two imaginary viruses.

There are two major lines of investigations that can illuminate the applicability of the CEP to viral infections in general. These include (i) the experimental evidence on the negative interactions between viruses, and (ii) the computational models of co-infection. In relation to the first line of research, a recent paper [10] employed a population-level approach, based on 44230 cases over 9 years, to track the epidemiological interactions between 11 strains of respiratory viruses, including influenza A and B, rhinoviruses, and three human coronaviruses (229E, NL63, HKU1). They inferred negative interaction between Rhinoviruses (A–C) and Influenza A virus—at both population and host levels—and negative interaction between Influenza A virus and Influenza B virus—at the population level. In addition, the three human coronaviruses showed negative interactions with Rhinoviruses (A–C), human Respiroviruses 1 and 4 at the population level.

While the aforementioned approach is a rarity in studying an organism, models that describe the co-infection of two viruses in a single host have gained growing attention over the past few years (e.g. [11, 12]). Gonzalez et al. [11] have used a mouse model to show how infection from rhinovirus strain 1B—before exposure to influenza A virus—reduces the severity of influenza in mice. They observed the same effect on the outcome of influenza when initially exposed to the hepatitis virus strain-1. Also recently, a more sophisticated experimental method using single-cell RNA localization has described how the entry of an influenza A virus restricts the replication of a second viral genome in the same cell [13]. This growing body of research highlights the prevalence of interaction between viruses in the same host; though many aspects of co-infection dynamics have yet to be experimentally elucidated.

The computational models of co-infection have primarily been concerned with the co-infection of a cell by multiple copies of the same virus, where these copies compete over target cells and resources (e.g., [14]). Here, we propose a stochastic-based model to explore the possible effects of co-infection by two species of virus. Although deterministic mathematical methods based on ordinary differential equations (ODE) are typical approaches to study the dynamics of viral infection, they often ignore the stochastic nature of the viral growth process. It has been shown that they can sometimes lead to inaccurate results because they mainly rely on the average reactions of viral infections [1519]). To simulate the stochastic multi-stage nature of the viral growth process occurring in viral co-infection, Gillespie algorithm, also known as the stochastic simulation algorithm, is usually considered as a trustworthy alternative. In particular, it supports simulating both discrete and stochastic behaviors of each stage of viral growth. The algorithm explicitly simulates the behavior of each virus in each stage of its growth process. Finally, it can provide a flexible framework to study the stochastic effects induced during the evolution of the infection. One characteristic of our simulation model is that several Gillespie algorithms run in infected cells simultaneously. This characteristic can represent what we presume of viral infection in one tissue. Gillespie’s algorithms are indeed independent of their features. To illustrate the co-infection of two viruses using Gillespie’s algorithm [20], we look at the dynamics of co-infection by introducing a benign and a malignant virus strains into a lattice of virtual cells. We investigate how the delay between the infection of the tissue by each strain, henceforth referred to as “co-infection delay” (D), affects the competition between the two strains.

The proposed model—called Co-Wish—offers a more in-depth study of the interaction between different virus strains competing in a shared or overlapping niche. The model could eventually be used to design strategies to induce immunity in the population against a deadly virus by introducing a competing but less-deadly strain into the population. The results suggest that when one virus has a slight advantage—e.g., a virus with a shorter infection cycle—over another, the one with the advantage will eventually dominate the other.

Results

To model the viral infection of tissue, we applied Gillespie’s stochastic algorithm [20] to simulate the spread of infection over the lattice of virtual cells. In our model—Co-Wish—viral infection is a five-stage process including attachment, penetration, eclipse, replication, and release. We treat each step of infection as a probabilistic process (Fig 1).

Fig 1. The four stages of the viral infection of a cell in the model.

Fig 1

The virus can move from stage i to stage j according to the transition probability Pij.

The viral growth is simulated by considering the duration of each stage (waiting time) and the probability of transition from one stage into the next (transition probability). The waiting times follow either an exponential, Exp(λ), or a Weibull distribution, W(λ, ω). All transition probabilities are generated from Beta distribution. The parameter values of the Beta distribution depend on the chosen distribution for the waiting time of the current stage. In the Methods section, we provide a detailed explanation of each of these steps in our model. In addition, we simulate the effect of the immune system on the dynamics of viral infection and how the interaction between virus strains affects their spread.

The dynamics of single strain infection

Before investigating the dynamics of co-infection in our model, we assessed the effects of different parameters of our model when a single strain infects the tissue (Fig 2). We ran the model in two different conditions. At first, the viral infection was simulated by setting the parameters of the models as default values(all the parameters were set as one.). Then, we changed the parameters due to shortening the waiting times.

Fig 2. Changing the waiting times for the different stages of the infection cycle influences the overall dynamic.

Fig 2

In our models, by default generate waiting times from Exponential distribution with one value for its parameter(Exp(λ = 1)), except for the eclipse duration which is generated from Weibull distribution. The scale and shape parameters of the Weibull distribution, both have been set as one(scale = 1, shape = 1). The blue trajectories show the behaviour of our model with the default parameters. To show how changing these parameters affects our model, in each of the four panels, we changed the waiting time of one of the transition (orange trajectories). For the manipulated simulations, the waiting times for transition between attachment, penetration, and replication were generated from Exp(λ = 3). The waiting time of the eclipse stage was generated from Weibull(1, 0.5). Each trajectory is the average of 100 runs. The shaded areas contain the trajectories of all the 100 runs for the given parameters. The immune system was not active in all the runs.

The results indicate that the shorter waiting times for different stages of the virus growth results in more viruses produced per infected cell and an increase in the number of infected cells. The impact of decreasing the waiting time on the production of new virions is more perceptible in the eclipse stage (Fig 2—Eclipse to Replication). Decreasing the waiting time at the eclipse stage increases the number of virions. Furthermore, it seems that the duration of the attachment stage directly affects the numbers of virions produced during the infection cycle (Fig 2—Attachment to Penetration)—perhaps because of a faster transition into the penetration stage allows the virus to replicate more virions in the host cell.

To visualize the viral infection of tissue, Co-Wish simulates the tissue as a lattice in which each node represents one cell. The dimension of the lattice is determined by the user. In our simulations, the host cell division is not designed. In addition, there is no competition between host cells in Co-Wish. Each infected cell can infect its neighboring cells. Co-Wish uses two patterns for determining the susceptible cells; quincunx and square. In the first pattern, an infected cell sits at the center of a quincunx and infects the four neighboring cells in the quincunx (Fig 3—Cross Pattern). In the second pattern, the infected cell is in the middle of a square that also includes the eight susceptible cells around it (Fig 3—Square Pattern). Co-Wish assumes each infected cell is able to infect at most three layers of cells in its vicinity. The proportions of the number of virions in each layer of infected cell vicinity are determined based on a mathematical equation, described in the Methods section.

Fig 3. Viral infection spreading patterns.

Fig 3

In our model, infection can either spread in cross (or quincunx) pattern or a square pattern. The red node demonstrates the infected cell. R determines how many layers of neighbouring cells can be affected by the infected cell in the middle of each pattern.

To select the pattern, we ran different simulations with both quincux and square patterns. The results showed that the two patterns of distributions only affect the pace at which the dynamics of infection unfold and not the observed patterns (Fig 4). Consequently, we used the square pattern that concludes faster dynamics to generate the results.

Fig 4. The patterns of the spread of the infection only affects the pace at which the dynamic of our model unfolds.

Fig 4

Each trajectory is the average of 100 runs. In all the runs, parameters of probability distributions to generate the waiting times were set as one. The shaded areas contain the trajectories of all the 100 runs for the given parameters. The immune system was not active in all the runs.

The influence of the immune system

In Co-Wish, the immune system response is characterized by the probability at which the immune system can eliminate the toxic elements (γ), the delay between the infection and the immune response, and the capacity of the immune system (I). Toxic elements include one or both of the following: infected cells; virions. In other words, Co-Wish can eliminate the infected cells, the virions, as well as the combination of both. Here, the capacity refers to the maximum number of each kind of toxic element that can be eliminated by the immune system. The capacity value has to be less than the number of toxic elements that are potentially produced in the early steps of the simulation. γ is drawn from a uniform distribution, U(a;b), that its parameters are within [0;1]. The immune system can eliminate toxic elements with a different probability. This characteristic indicates that all toxic elements produced in each step of the simulation do not have the same probability of elimination. For example, Co-Wish generates separate probability values for the killing of each virion. Here, we assume that a virion is killed with probability p and survives with probability 1 − p.

After the infection, the immune system mounts a response with a short delay, thus a number of cells become infected before the immune system attempts to combat the infection. The immune system as described here does not distinguish between the innate immune response and the adaptive response, but in general terms, it resembles the innate response since it lacks memory of previous encounters with any of the two strains in our simulations and responds to both pathogens in equal measures.

In our simulations, we employed four broad types of immune response:

  1. Weak with limited capacity (WL): γU(0, 0.4)

  2. Weak with unlimited capacity (WU):γU(0, 0.4)

  3. Strong with limited capacity (SL): γU(0.6, 1)

  4. Strong with unlimited capacity (SU): γU(0.6, 1)

Limited capacity indicates that the immune system will not be active in all the steps of the simulation. It will be deactivated when the capacity hits the predetermined value for I. If the capacity is unlimited, the immune system will be active by the end of simulation.

Different levels of immune system strength have been studied in a single virus model. Fig 5 represents the number of virions for four levels of the immune system. As expected, decrease the killing rate and putting the restriction to the capacity of the immune system, decrease its strength as well and the slopes of the graphs are steeper for the weaker immune systems.

Fig 5. The number of virions produced during the infection is affected by the strength of the immune system.

Fig 5

We utilised four different modes of immunity in our model: weak and limited capacity (WL), weak and unlimited capacity (WU), strong and limited capacity (SL), strong and unlimited capacity (SU). Each trajectory is the average of 100 runs. In all the runs, parameters of probability distributions to generate the waiting times were set as one. The shaded areas contain the trajectories of all the 100 runs for the given parameters.

Competition during co-infection

In this study, we assumed that the benign strain (virus B) has a shorter infection cycle than the malignant strain (virus M). Therefore, we set the parameters of the model such that the virus B spends less time in each of the four stages of viral growth—from attachment to replication—compared with the virus M. In addition, we assumed that the burst size of virus B is less than virus M, due to its shorter waiting time in the replication stage. The location of the second virus in the lattice is chosen randomly, and it is independent of the first virus location. Those cells that are not infected by the first virus can be a niche of the second virus.

To simulate the co-infection, we considered three scenarios:

  1. Two viruses infect the host simultaneously.

  2. The host is infected by B and later by M

  3. The reverse of second scenario

Co-infection delay D refers to the time delay between the introduction of each viral strain to the tissue. In the first scenario, the co-infection delay is zero. If the co-infection delay is set as 30 steps, it means the second virus enters into the model thirty steps after the first virus. During this time, the first virus starts to infect some cells. In Co-Wish, only one strain enters a cell—thus, cells are not co-infected—but the pattern of the spread of each strain in a co-infected fashion affects the spread of the strain by limiting the susceptible cells in the tissue.

To test the three scenarios of co-infections in our model—namely, (i) no delay between the introduction of M and B strains, (ii) B is introduced first, and (iii) M is introduced first—we simulated each scenario (using parameters in Table 1). The co-infection delay (D) in each scenario, was set to 0, + 30, and −30 steps, respectively. In the absence of the immune response, the benign strain is always dominant (Fig 6). The results suggest that the waiting times greatly influence the dynamics of co-infection. While the B strain spends less time in the replication stage and, consequently, produces a smaller burst compared to the M stain, its shorter infection cycle enables it to out-compete the M stain during co-infection.

Table 1.

(*): c = 100; The parameter p has been calculated according to Eqs (1) and (2); to calculate P1, the time T has been set as 5. (**):P1 has been calculated according to Eqs (4) and (5). The parameters β0 and β1 in Eq (4) have been set as zero and 1, respectively.

virus Viral Growth Stage Waiting Time Distribution Transition Probability Distribution
B Attachment-Penetration exp(λ);λ > 1 Beta(1,1)
Penetration-Eclipse exp(λ);λ > 1 Beta(1, 1)
Eclipse-Replication weibull(λ = 1, ω);ω < 1 Beta(cp, c(1 − p))*
Replication-Shedding exp(λ);λ > 1 P1**
M Attachment-Penetration exp(λ = 1) Beta(1, 1)
Penetration-Eclipse exp(λ = 1) Beta(1, 1)
Eclipse-Replication weibull(λ = 1, ω = 1) Beta(1, 1)
Replication-Shedding exp(λ = 1) P1**

Fig 6. Benign strain is dominant in all three conditions of co-infection delay (D).

Fig 6

Introduction of the benign strain (B) 30 steps before the introduction of the malignant strain (D=+30), or simultaneously (D=0) greatly hampers the spread of B strain but introducing this strain 30 steps after the introduction of M strain (D=-30) tighten the competition between the strains. Each trajectory is the average of 100 runs. The shaded areas contain the trajectories of all the 100 runs for the given parameters. For virus M, all the parameters of probability distributions to generate the waiting times were set as one. For virus B, the waiting times had been generated from Exp(λ = 3)and Weibull(λ = 1, ω = 0.5). The immune system was not active in all the runs.

Since the main idea of this work is to explore the theoretical feasibility of viral co-infection of two malignant and benign strains to reduce the spread of malignant one, we simulated the third scenario—B is introduced first—separately by changing the parameters of applied distributions to generate the waiting times. We considered four conditions. In each condition, the waiting times have been shortened by increasing the value of Exponential distribution parameters and decreasing the shape parameters of Weibull distribution. In other words, we investigated how changing the waiting times can influence the number of steps at which the B strain out-competes the M strain (Fig 7). The results show that reducing the waiting times decreases the time it takes for the B strain to out-compete the M strain (Fig 8).

Fig 7. Changing the parameters of probability distributions to generate waiting times affects the number of overtaking step of benign strain when co-infection delay is (-30).

Fig 7

The parameters include λ, the parameter of Exponential distribution and ω, the parameter of Weibull distribution. Condition 1: λ = 2, ω = 1/2; condition 2: λ = 3, ω = 1/3; condition 3: λ = 4, ω = 1/4; condition 4: λ = 5, ω = 1/5. Boxes show the median and first and third quartiles and the notches represent the %95 confidence interval around the median. The immune system was not active in all the runs.

Fig 8. Different waiting times expedites benign strain (B) to outcompete malignant strain (M) when co-infection delay is (-30).

Fig 8

To generate different waiting times for virus B growth process, λ–the parameter of exponential distributions–and ω–the shape parameter of the Weibull distributions–have been changed. Condition 1: λ = 2, ω = 1/2; condition 2: λ = 3, ω = 1/3; condition 3: λ = 4, ω = 1/4; condition 4: λ = 5, ω = 1/5. For virus M, all the parameters of probability distributions to generate the waiting times were set as one. The immune system was not active in all the runs.

Discussion

In this study, we look into the dynamics of viral co-infection using a CEP-inspired computational model. To explore the possible effects of co-infection by two virus species, we develop a stochastic-based viral infection simulator—called Co-Wish—to model and display the competition between two strains. Co-Wish offers a flexible framework to study the stochastic effects induced during the evolution of the infection. Co-Wish supports simulating both discrete and stochastic behaviors of each stage of viral growth. It explicitly simulates the behavior of each virus in each stage of its growth process. Co-Wish also incorporates the effect of the innate immune system on the dynamics of viral infection.

Several studies consider the concept of CEP to illustrate the competition between viruses [2123]. In particular, Peter and colleagues in [24] recently presented a survey on SARS-CoV-2 infection dynamics models, focusing mostly on ODE-based methods. Typical models of virus dynamics are developed based on three ordinary differential equations concentrating on the number of target cells, free virions, and infected cells [15]; i.e., most of these approaches proposed for mathematical modeling of host-pathogen interactions can be seen as an extension of the ODE-based model [2527]. Although these extended models provide explainable frameworks in which they improve our understanding of the dynamic of the viral infection within the host, they somehow suffer the main issue about deterministic models, namely, relying on viruses’ average behaviors during the infection process [16, 17]. As a result, ODE-based approaches often ignore the stochastic nature of the viral infection. Therefore, controlling the heterogeneity of virus populations affecting competitive exclusion is not readily achievable in the ODE-based methods. Accounting for the viral infections stochastic nature during the simulation process, in contrast, is one of Co-Wish’s most essential features, enabling our model to demonstrate the competition between virus strains more realistically.

Using Co-Wish, we first simulate the characteristics of two virus growth processes separately. Then, we investigate their interactions until one can dominate the other. To model the co-infection of two viruses using Co-Wish, we consider three scenarios including (i) two viruses infecting the host at the same time,(ii) delayed initiation of the malignant-virus infection, and (iii) delayed initiation of the benign-virus infection. While an earlier stage of infection initiation for one of the viruses can be considered a competitive advantage, it plays a minor role as compared to the waiting times in the five stages of the viral infection of a cell in the proposed model—namely attachment, penetration, eclipse, replication, and release. In summary, we observed that these waiting times in the five stages of the viral infection—regardless of varying the starting times for the viruses—crucially affect the competition between two viruses over a shared host. In other words, if the starting time for the benign virus with a higher viral growth rate is not delayed too long, it can eventually block the infection caused by the malignant virus. In particular, our results showed that the waiting times at the eclipse stage significantly affected the number of virions in the competition between the two viruses. We also observed a similar trend when we incorporated an innate immune response during viral co-infection.

The dynamics of viral co-infection remain one of the lesser-explored areas of virology and epidemiology. The model presented in this paper illustrates how the competition between viral strains over a shared niche can hinder the propagation of one, an unsurprising result given the competitive exclusion principle. While the model is inescapably a simplified reconstruction of what might happen in nature, a few lines of experimental studies discussed below provide some credence to the hope that the dynamic of co-infection can be employed to control the spread of lethal strain by introducing a more benign strain, but many obstacles must be overcome before this hope can become a viable epidemiological strategy.

Previous studies have investigated the association between co-infection with the GB virus C and survival among individuals with HIV infection [28, 29]. For example, George and colleagues reported a significantly higher survival rate among HIV-positive patients co-infected with the GB virus C virus than among HIV-positive individuals without GB virus C co-infection [28]. This result—a beneficial effect of co-infection with the GB virus C virus on HIV-related survival—suggested the GB virus C infection as a potential treatment for HIV-infected patients [28]. Our CEP-based model also suggests that one viral strain might be able to out-compete and eliminate a rival strain in the same ecological niche during a co-infection scenario. Suggestively, our theoretical framework hints at an intriguing research direction in the dynamics of co-infection, viz. the possibility of controlling the spread of any viral infection in the population by supplanting the lethal virus with a much less fatal one. For example, researchers could employ our theoretical model to explore and evaluate the CEP-inspired hypothesis in tackling a viral respiratory pandemic, such as COVID-19, through co-infection with some forms of common respiratory viruses that are less harmful. However, in the case of COVID-19, it could prove challenging to find such a harmless virus that would keep SARS-Cov-2 below the exposure level. Also, there is no guarantee that—even if our proposed CEP-based hypothesis works—SARS-Cov-2 does not come up as the winner of such competition. Nevertheless, there are four harmless human Coronaviruses—namely HCoV-NL63, HCoV-229E, HCoV-HKU1, and HCoV-OC43—that lead to a range of respiratory diseases with mild symptoms—cause approximately 15% of common colds [30]. Recent studies [31, 32] also have revealed that hospitalized COVID-19 patients who developed severe symptoms of COVID-19, such as high fever and pneumonia, were rarely co-infected with other common respiratory viruses, particularly the common colds. From the immune system’s point of view, this might be related to the result of a new study [33] that discusses that COVID-19 patients previously infected with other forms of the human beta-coronaviruses could also develop some types of immunity to SARS-CoV-2. Yet, further experimental investigations are needed to understand the therapeutic potentials of viral co-infection thoroughly. In particular, there is an urgent need to establish clinical trials assessing the role of simultaneous respiratory viral infection in the variability of disease virulence among COVID-19 patients.

It is also worth mentioning that the notion that the dynamics of viral co-infection in respiratory diseases can be dominated by competition was previously highlighted by Burattini and colleagues [21] and Pinky and Dobrovolny [19]. The former resorts to CEP as a premise to investigate the conditions under which two or more strains can coexist within a host, while in the latter, Pinky and Dobrovolny explored the possible competition between influenza A virus and a handful of respiratory viruses [19]. Although their premise is the aforementioned studies start from a postulation similar to our work, their approach—in modeling the co-infection dynamics using ODEs—is quite different from the one applied in this work. In particular, we modeled the infection at the cell level. Also, our model—as opposed to the one proposed by Burattini and colleagues and Pinky and Dobrovolny—includes the immune response Neither of these main differences is expected to profoundly change the qualitative behavior of two viral strains competing over an ecological niche. Nevertheless, we believe that the added complexity of the model is a step in the right direction. Thus, the present study can be considered the logical extension of previous works on this topic.

There are several important limitations to the idea behind our approach. Crucially, it would be deeply preposterous to suggest a model of viral co-infection as a basis for prescribing remedies for public health issues, such as the COVID-19 pandemic. Anyone with a cursory knowledge of biology, specifically epidemiology and ecology, understands the limitations of even the best and most detailed models in predicting the interaction between living entities. Rather, introducing the notion of Gause’s law—a quintessentially ecological concept—to the dynamics of co-infection is one of the first steps to explore the therapeutic ramifications of the CEP as applied to viral co-infection. The journey from this simple theoretical curiosity to a usable remedy against viral epidemics undoubtedly will have to be paved by a thick layer of theoretical and experimental studies. Given that many characteristics of viral life cycles remain elusive (e.g., the burst size of different viral strains), applying Gause’s law to combat viral epidemics will likely remain a fanciful notion for many years to come.

Methods

Co-Wish—the proposed model—is a stochastic-based model to simulate the stages of viral infection. To model the viral infection, we consider each stage of infection as a probabilistic process. Each stage is simulated by applying Gillespie’s stochastic algorithm [20].

Viral growth process

To simulate the viral growth process, Co-Wish considers the duration of each state (waiting time) and the probability of transition from one state to the next one (transition probability) The waiting times follow either an exponential, Exp(λ), or a Weibull distribution, W(λ, ω). Weibull distribution is a widely used distribution to generate time-to-failure data. If variable t is defined as time-to-failure, the Weibull distribution describes the hazard rates as being proportional to a power of time. It is worth pointing out that in survival analysis, the event of interest is typically referred to as a ‘Failure.’ The hazard rate is the potentiality that an event will occur within an immeasurably short time interval (between time t and t + Δt, as Δt goes to 0), given that the event has not yet occurred—i.e., the instantaneous risk that an event will happen to a particular patient at a certain time, given that the patient at risk has survived up to that time. In the Weibull distribution, the hazard function is h(t) = λωtω−1, where λ > 0 and ω > 0. The parameter ω —called a shape parameter— determines the shape of the hazard function. If ω > 1, then the hazard increases as time increases. If ω = 1, then the Weibull distribution collapses to an Exponential distribution in which the hazard function is constant (h(t) = λ). If ω < 1, then the hazard rate decreases over time [34]. Applying the Weibull distribution allows us to control the hazard rate of transition to the next stage by determining different values for the shape parameter ω.

All transition probabilities are generated from the Beta distribution. The parameter values of the Beta distribution depend on the chosen distribution for the waiting time of the current stage. If the waiting times of one stage are generated from an Exponential distribution, the corresponding transition probabilities will be generated from Beta(1, 1). If waiting times are drawn from a Weibull distribution—inspired by the survival analysis concepts—the transition probabilities will be generated based on the probability of remaining in the current stage [34]).

The survival function for the Weibull distribution determines probability of remaining in a given stage [34], i.e., the transition into the next stage is our event of interest and the time T is defined as the maximum time it takes for a virus to transmit into the next stage. Then, we use the survival function (S(T)) for Weibull distribution to determine the probability of transition from the current stage to the next one P:

S(T)=exp(-λTω),P=1-S(T). (1)

We consider P as the expected value of the Beta distribution (E(t)) and determine the parameters of Beta distribution according to the value of P:

E(t)=αα+β=P (2)

To estimate the parameters, we keep the sum of the parameters constant, i.e., α + β = C. Consequently, the parameters of the beta distribution will be

α=CP,β=C(1-P). (3)

In addition, to avoid getting zero values for the parameters, when P is either zero or one, a positive small value (ϵ) is added to the above estimation.

To determine the probability of transition from the replication stage to the release stage, we assume that the probability is a function of the number of viruses produced during the replication stage. In other words, We consider the transition of replication stage to the release stage as a binary event. We assume a linear relationship between the number of produced viruses in the cell(x) and the log-odds of transition from the replication stage to the shedding stage:

logbprs1-prs=β0+β1x, (4)

where prs is transition probability from replication stage into the shedding stage. Therefore,

prs=bβ0+β1x1+bβ0+β1x. (5)

We suppose increasing the number of produced virions in the replication stage, increases the log-odds of transitioning to the release stage. Therefore, the positive value should be determined for parameter β1. A positive value for β1 means that increasing x by one increase the log-odds by β1. The number of viruses produced in a cell depends on the waiting time of the replication stage, i.e., longer waiting times results in the production of more viruses. During the replication stage, the virions are replicated according to stochastic lags. These lags between replication of each virus is determined stochastically. The replication process continues until these lags add up to the total waiting time of the replication stage.

The spread of viral infection

After determining the patterns of infection, we approach to define the proportion of number of virions attached to each susceptible cell in three layers in its vicinity (Vi;i = 1, 2, 3):

Vi=kiVni, (6)

where V is the number of virions released from an infected cell and ni is the number of susceptible cells in each layer of cells surrounding the infected cell, according to either cross or square pattern. In the square pattern, n1, n2, and n3 are 8,16, and 24, respectively. ki is a weight designating what shares of produced virions belong to each layer of vicinity. For example, if the total number of released virions from one cell is 1000 and k1 = 0.4, k2 = 0.3, and k3 = 0.3, 400, 300, and 300 virions will be belonged to each layer, respectively. In each layer, the virions are diffused uniformly among cells.

Infection calculation: Number of virions and infected cells

The number of infected cells and free virions is calculated in each step of the simulation. The time length of each step is predefined before starting the simulation. Hence, the time length of the steps is constant. In our study, the time length of the simulation steps is about 1000 milliseconds. Co-Wish runs a number of Gillespie algorithms simultaneously. In other words, many cells can be infected at the same time.The time lengths of the algorithms with different settings are not alike. Each step is terminated when the sum of the time length of different runs of Gillespie algorithm is equal to the determined time length of the step.

Supporting information

S1 File

(PDF)

Data Availability

The program written simulate the spread of the competition between viruses is available at https://github.com/safarvafadar/virusinfection.

Funding Statement

The authors received no specific funding for this work.

References

  • 1. Hardin G. The Competitive Exclusion Principle. Science. 1960;131(3409):1292–1297. 10.1126/science.131.3409.1292 [DOI] [PubMed] [Google Scholar]
  • 2. Darwin CR. On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. 1st ed John Murray; 1859. [PMC free article] [PubMed] [Google Scholar]
  • 3. Kakizoe Y, Nakaoka S, Beauchemin CAA, Morita S, Mori H, Igarashi T, et al. A method to determine the duration of the eclipse phase for in vitro infection with a highly pathogenic SHIV strain. Scientific reports. 2015;5:10371–10371. 10.1038/srep10371 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Schneider M, Ackermann K, Stuart M, Wex C, Protzer U, Schätzl HM, et al. Severe Acute Respiratory Syndrome Coronavirus Replication Is Severely Impaired by MG132 due to Proteasome-Independent Inhibition of M-Calpain. Journal of Virology. 2012;86(18):10112–10122. 10.1128/JVI.01001-12 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Bar-On YM, Flamholz A, Phillips R, Milo R. SARS-CoV-2 (COVID-19) by the numbers. eLife. 2020;9:e57309 10.7554/eLife.57309 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Flint J, Racaniello VR, Rall GF, Skalka AM, Enquist LW. Principles of Virology. AMS Press; 2015. [Google Scholar]
  • 7. Rainey PB, Travisano M. Adaptive radiation in a heterogeneous environment. Nature. 1998;394(6688):69–72. 10.1038/27900 [DOI] [PubMed] [Google Scholar]
  • 8. Lauer SA, Grantz KH, Bi Q, Jones FK, Zheng Q, Meredith HR, et al. The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application. Annals of Internal Medicine. 2020. 10.7326/M20-0504 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Jiang X, Rayner S, Luo MH. Does SARS-CoV-2 has a longer incubation period than SARS and MERS? Journal of medical virology. 2020;92(5):476–478. 10.1002/jmv.25708 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Nickbakhsh S, Mair C, Matthews L, Reeve R, Johnson PCD, Thorburn F, et al. Virus-virus interactions impact the population dynamics of influenza and the common cold. Proceedings of the National Academy of Sciences of the United States of America. 2019;116(52):27142–27150. 10.1073/pnas.1911083116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Gonzalez AJ, Ijezie EC, Balemba OB, Miura TA. Attenuation of Influenza A Virus Disease Severity by Viral Coinfection in a Mouse Model. Journal of virology. 2018;92(23):e00881–18. 10.1128/JVI.00881-18 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Goto H, Ihira H, Morishita K, Tsuchiya M, Ohta K, Yumine N, et al. Enhanced growth of influenza A virus by coinfection with human parainfluenza virus type 2. Medical microbiology and immunology. 2016;205(3):209–218. 10.1007/s00430-015-0441-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Dou D, Hernández-Neuta I, Wang H, Östbye H, Qian X, Thiele S, et al. Analysis of IAV Replication and Co-infection Dynamics by a Versatile RNA Viral Genome Labeling Method. Cell Reports. 2017;20(1):251–263. 10.1016/j.celrep.2017.06.021 [DOI] [PubMed] [Google Scholar]
  • 14. Phan D, Wodarz D. Modeling multiple infection of cells by viruses: Challenges and insights. Mathematical biosciences. 2015;264:21–28. 10.1016/j.mbs.2015.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Perelson AS. Modelling viral and immune system dynamics. Nature Reviews Immunology. 2002;2:28–36. 10.1038/nri700 [DOI] [PubMed] [Google Scholar]
  • 16. Srivastava R, You L, Summers J, J Y. Stochastic vs. Deterministic Modeling of Intracellular Viral Kinetics. Journal of Theoretical Biology. 2002;218(3):309–321. 10.1006/jtbi.2002.3078 [DOI] [PubMed] [Google Scholar]
  • 17. Yuan Y, Allen LJS. Stochastic models for virus and immune system dynamics. Mathematical Biosciences. 2011;234:84–94. 10.1016/j.mbs.2011.08.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Iwami S, Sato K, De Boer R, Aihara K, Miura T, Koyanagi Y. Identifying viral parameters from in vitro cell cultures. Frontiers in Microbiology. 2012;3:319 10.3389/fmicb.2012.00319 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Pinky L, Dobrovolny HM. Coinfections of the Respiratory Tract: Viral Competition for Resources. PLOS ONE. 2016;11(5):e0155589–. 10.1371/journal.pone.0155589 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Gillespie DT. Stochastic Simulation of Chemical Kinetics. Annual Review of Physical Chemistry. 2007;58(1):35–55. 10.1146/annurev.physchem.58.032806.104637 [DOI] [PubMed] [Google Scholar]
  • 21. Burattini M, Coutinho F, Massad E. Viral evolution and the competitive exclusion principle. Bioscience Hypotheses. 2008;1(3):168–171. 10.1016/j.bihy.2008.05.003 [DOI] [Google Scholar]
  • 22. Dai L, Zou X. Analysis of a within-host age-structured model with mutations between two viral strains. Journal of Mathematical Analysis and Applications. 2015;426(2):953–970. 10.1016/j.jmaa.2015.01.032 [DOI] [Google Scholar]
  • 23. Tang X, Wang Z, Yang J. Threshold dynamics and competitive exclusion in a virus infection model with general incidence function and density-dependent diffusion. Complexity. 2020;2020 10.1155/2020/4923856 [DOI] [Google Scholar]
  • 24. Peter S, Dittrich P, Ibrahim B. Structure and Hierarchy of SARS-CoV-2 Infection Dynamics Models Revealed by Reaction Network Analysis. Viruses. 2021;13(1):14 10.3390/v13010014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Amaku M, Burattini MN, Coutinho FAB, Massad E. Modeling the competition between viruses in a complex plant–pathogen system. Phytopathology. 2010;100(10):1042–1047. 10.1094/PHYTO-10-09-0289 [DOI] [PubMed] [Google Scholar]
  • 26. Ewald J, Sieber P, Garde R, Lang SN, Schuster S, Ibrahim B. Trends in mathematical modeling of host–pathogen interactions. Cellular and Molecular Life Sciences. 2020;77(3):467–480. 10.1007/s00018-019-03382-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Koelle K, Farrell AP, Brooke CB, Ke R. Within-host infectious disease models accommodating cellular coinfection, with an application to influenza. Virus evolution. 2019;5(2):vez018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Xiang J, Wünschmann S, Diekema DJ, Klinzman D, Patrick KD, George SL, et al. Effect of Coinfection with GB Virus C on Survival among Patients with HIV Infection. New England Journal of Medicine. 2001;345(10):707–714. 10.1056/NEJMoa003364 [DOI] [PubMed] [Google Scholar]
  • 29. George SL, Wünschmann S, McCoy J, Xiang J, Stapleton JT. Interactions between GB virus type C and HIV. Current Infectious Disease Reports. 2002;4(6):550–558. 10.1007/s11908-002-0044-9 [DOI] [PubMed] [Google Scholar]
  • 30. Fehr AR, Perlman S. Coronaviruses: an overview of their replication and pathogenesis. Methods in molecular biology (Clifton, NJ). 2015;1282:1–23. 10.1007/978-1-4939-2438-7_1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Nowak MD, Sordillo EM, Gitman MR, Paniz Mondolfi AE. Co-infection in SARS-CoV-2 infected Patients: Where Are Influenza Virus and Rhinovirus/Enterovirus? J Med Virol. 2020. 10.1002/jmv.25953 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Blasco ML, Buesa J, Colomina J, Forner MJ, Galindo MJ, Navarro J, et al. Co-detection of respiratory pathogens in patients hospitalized with Coronavirus viral disease-2019 pneumonia. J Med Virol. 2020. 10.1002/jmv.25922 [DOI] [PubMed] [Google Scholar]
  • 33. Bert NL, Tan AT, Kunasegaran K, Tham CYL, Hafezi M, Chia A, et al. Different pattern of pre-existing SARS-COV-2 specific T cell immunity in SARS-recovered and uninfected individuals. bioRxiv. 2020. [Google Scholar]
  • 34. Kleinbaum DG, Klein M. Survival Analysis. 3rd ed New York: Springer-Verlag; 2012. [Google Scholar]

Decision Letter 0

Bashar Ibrahim

23 Sep 2020

PONE-D-20-22300

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: a computational perspective

PLOS ONE

Dear Dr. Sadeghi,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Nov 07 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

  • A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

  • An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Bashar Ibrahim

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: N/A

Reviewer #3: N/A

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: No

**********

5. Review Comments to the Author

Reviewer #1: Dear authors,

please notice the following comments and suggestions regarding your paper and correct the regarding parts if necessary:

(1) lines 101 - 103: Is there a word missing?

(2) caption of Fig. 2: "The each trajectory..." This sentence does not make sense. It appears several times in the paper !!!

(3) !!! Almost all references to the Figure numbers in the text seem to be wrong, i.e. shifted by one !!!

(4) line 178: "strong With limited capacity" - "w" is missing

(5) lines 224 - 226: The plot does not really look LINEAR. How do you draw this conclusion. It could also be a quadratic or exponential decline.

Reviewer #2: This study tries to assess the feasibility of use of co-infecting virus to combat problematic virus, based on simulation results. The main conclusion obtained by the authors is that if co-infecting (harmless) virus can accumulate more rapidly than the target (harmful) virus and if the two viruses compete for the same type of host cells, the target virus will be dominated by the co-infecting virus, even when the co-infecting virus came later; to be honest, this conclusion will not surprise anyone, but I can find a potential value of this study in establishment of a framework for discussion on “under what circumstances such antiviral strategy can function.” To fulfill the potential value of the study, I recommend the authors to consider the following points:

Major points:

1) The simulation results are totally dependent on parameter values, especially on lambda. Please provide lambda (and other parameter) values used for the simulations explicitly in every figure; to avoid confusion, clearly define parameters. For example, define different lambda value for each virus by using lambda B and lambda M.

2) Please provide average waiting times for each step, as readers can compare them with D.

3) Figures and figure legends do not correspond to each other in some instances. I need explanation for Fig. 3 but I wonder where I can find it...

4) In Fig. 6, B virus continues to accumulate but M virus does not. Could the authors provide intuitive explanation of this phenomenon? For example, limitation in the number of host cells will not explain because B virus would also suffer.

5) In relation to the above comment, I’m not sure if the authors assumed (1) limitation in the number of cells, (2) host cell division, and (3) innate immunity in each figure.

6) In my understanding, authors think that competition for host cells is important in the simulation. Could the authors provide the time-course change of the number (and/or proportion) of cells attacked by both of the virus (before they get infected)? I’m not sure if the local spread of viruses assumed in this manuscript allows competition for host cells, though it totally depends on the virus and host cell densities assumed.

7) L40 and L312: the authors touch incubation times of viruses but it was not clear to me how they are relevant to this study.

Other minor points:

8) Figure 2 Legend: Replication -> replication

9) L178: ith -> with

10) The reference formatting is terrible.

Reviewer #3: In this paper the authors have presented an interesting framework for the progression of viral infection in host tissue. The model simulates the invasion of two competing viruses in a theoretical lattice of initially susceptible cells. Cells that are infected by one virus are prevented by being simultaneously infected by a competing strain to simulate the effect of the competitive exclusion principle (CEP). The premise of this paper therefore may be appealing to not only mathematical modellers but also biologists and medics working in related fields.

However, I do have some key concerns, that if addressed would benefit the manuscript:

- The authors do not provide a strong justification for the application of their model to SARS-CoV-2 in their introduction, which detracts from the narrative since the reader is left questioning the relevance of the model in this setting. Some justification is provided later in the discussion section, however, this is not particularly strong, and thus the application of the model to SARS-CoV-2 feels very extrapolative. I would therefore encourage the authors to consider removing the emphasis of the model’s application to SARS-CoV-2 in the abstract and introduction of the paper.

- The action of the CEP in the model was buried in the paper (it was briefly alluded to in lines 201 -203). This could have been made much clearer. Moreover, it appears that the authors are more specifically modelling superinfection exclusion, whereby secondary infection is prevented at a cellular level, which I would consider to be a specific example of the CEP.

- The paper would benefit if the model was described in clearer detail. For example, the authors have not made the action of the immune response in the model clear. I appreciate that to appeal to a wider audience that the authors may not wish to go into a great level of detail in the main body of the text, but I feel a more detailed description of all of the components of the model in the supplementary information would be very beneficial.

- The paper would benefit from a more thorough justification of the parameterisation of the model, with suitable references to existing literature. While I appreciate different parameter combinations have been explored, this is somewhat limited, and I feel a more detailed sensitivity analysis would be beneficial. This also would most likely be appropriate to include in the supplementary information.

- The authors do not appear to use time units in their parametrisation of the model, or in the results presented, where all figures are presented in “steps” along the x-axis.

- The above point raises significant concerns – since the model is based on a Gillespie algorithm, each time step will be of a different time length due to different sojourn times between events. It could therefore be misleading to plot the trajectories of their results (e.g. number of virions) against the step count for each run of the simulation, since these are evolving over different timescales. I would therefore strongly advice that the model is revisited to ensure that all x-axes are presented over some units of time (hours or days).

- Additionally, a thorough spelling and grammar check is required

While I do have a significant number of concerns with the manuscript in its current form, I do believe that a valuable piece of work could result from the model framework that the authors have used. Moreover, I felt the premise of the model is interesting in principle and would like to thank the authors for sharing their work.

I have made some additional inline comments below:

line 69 - 70: If there is evidence of a reduced risk of a subsequent coronavirus infection by another respiratory virus then I think this needs to be explicitly stated to be able to back up the paper's justification for applying the model to SARS-CoV-2. Currently, it's not specified which pairs of viruses had a negative interaction between them in ref [10].

Line 133: The legend for figure 2 does not appear to match figure 2 in this document, which appears to show the different spreading patterns (square vs cross)

line 165: Does "the capacity of the immune system (I)" have any units, such as the maximum number of infected cells/virions eliminated per day? It is not clear how "I" is defined mathematically in the model.

line 167: Gamma appears to have been defined in lines 163-164 as "the rate at which the immune (system) eliminates infected cells or virions". Therefore, it presumably has units of "cells/virions cleared per hour/day". I therefore do not understand the logic behind why it is then drawn from a U[0,1] distribution.

Line 176: to differentiate this from the "unlimited immune response capacity" situation, I think it would be better to have notation along the lines of 0>i<i_c bound.="" i_c="" is="" some="" upper="" where="">

line 199: It is unclear how the second virus is seeded into the model. It would be helpful if the authors clarified this (e.g. is it seeded into a cell in the lattice at random?)

line 213: There is a lack of justification for the parameters chosen for the various distributions in table 1. While there may be limited literature for SARS-CoV-2, literature on other respiratory viruses may allow the authors to justify their choices. As highlighted before, I feel the paper would benefit if all times are given in appropriate units such as days, rather than being treated as dimensionless. I would also encourage the authors to consider including a sensitivity analysis of the different parameters in the supplementary information.

lines 267-271: Hepatitis G is now typically referred to as GB virus C

lines 312-314: If the authors wish to keep the focus of the paper on SARS-CoV-2, I feel the manuscript would benefit if these studies (refs [30] and [31]) were highlighted in the introduction section. The reader would then understand the authors justification for the model to be placed in the context of SARS-CoV-2.>/i<i_c>

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes: Andrew C Glover

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2021 Feb 24;16(2):e0247200. doi: 10.1371/journal.pone.0247200.r002

Author response to Decision Letter 0


30 Oct 2020

Point-by-Point responses to Reviewer #1

Dear authors,

Please notice the following comments and suggestions regarding your paper and correct the regarding parts if necessary:

We would like to thank the reviewer for the valuable comments and suggestions. We revised and improved our manuscript accordingly. The responses to the comments are as follows.

C1-1) lines 101 - 103: Is there a word missing?

A1-1) It is now edited.

C2-1) Caption of Fig. 2: "The each trajectory..." This sentence does not make sense. It appears several times in the paper!!!

A2-1) They are all edited now.

C3-1) Almost all references to the Figure numbers in the text seem to be wrong, i.e. shifted by one!!!

A3-1) We are sorry for the confusion we have caused. The figure numbers are now corrected.

C4-1) Line 178: "strong With limited capacity" - "w" is missing.

A4-1) It is now edited.

C5-1) Lines 224 - 226: The plot does not really look LINEAR. How do you draw this conclusion? It could also be a quadratic or exponential decline.

A5-1) The main idea of the plot is to display the effect of shortening the viral growth process, by generating shorter waiting times, on the time at which the benign virus out-competes the malignant virus. We did not study the statistical association which is possible between the value of the probability distribution parameters and the out-competing time. In the revised version of the manuscript, we have omitted the sentence which had made misunderstanding about the kind of correlation between probability distributions parameters and the benign virus overtaking time. The text in lines[225]-[234] now reads:

“Since the main idea of this work is to explore the theoretical feasibility of viral co-infection of two malignant and benign strains to reduce the spread of malignant one, we simulated the third scenario—B is introduced first—separately by changing the parameters of applied distributions to generate the waiting times. We considered four conditions. In each condition, the waiting times have been shortened by increasing the value of Exponential distribution parameters and decreasing the shape parameters of Weibull distribution. In other words, we investigated how changing the waiting times can influence the number of steps at which the B strain out-competes the M strain(Figure7). The results show that reducing the waiting times decreases the time it takes for the B strain to out-compete the M strain (Figure8).”

Point-by-Point responses to Reviewer #2

This study tries to assess the feasibility of use of co-infecting virus to combat problematic virus, based on simulation results. The main conclusion obtained by the authors is that if co-infecting (harmless) virus can accumulate more rapidly than the target (harmful) virus and if the two viruses compete for the same type of host cells, the target virus will be dominated by the co-infecting virus, even when the co-infecting virus came later; to be honest, this conclusion will not surprise anyone, but I can find a potential value of this study in establishment of a framework for discussion on “under what circumstances such antiviral strategy can function.” To fulfill the potential value of the study, I recommend the authors to consider the following points:

We thank the reviewer for her/his careful and critical reading of our manuscript and his/her valuable comments and suggestions. We revised and improved our manuscript accordingly. The responses to the comments are as follows. We revised and improved our manuscript accordingly.

MajorPoints:

C2-1) The simulation results are totally dependent on parameter values, especially on lambda. Please provide lambda (and other parameter) values used for the simulations explicitly in every figure; to avoid confusion, clearly define parameters. For example, define different lambda value for each virus by using lambda B and lambda M.

A2-1) The more explanation about parameters have been now included in the figures.

C2-2) Please provide average waiting times for each step, as readers can compare them with D.

A2-2) There are three different concepts related to time in our manuscript: Waiting time, simulation step and co-infection delay.

Waiting time is the duration of each stage of the viral growth process. Waiting times are generated from Exponential or Weibull distributions. The waiting times are scaled by common units of time such as milliseconds. For example, if the generated value for one waiting time is ten, we consider the time length of the corresponding stage as ten milliseconds. In the simulation model, different Gillespie algorithms—different viral growth processes—are run at the same time. Since, the time lengths of the algorithms are different, we assume the constant time length for the simulations step, such as L. One step of the simulation includes a number of Gillespie algorithms which the sum of their time lengths is L. The plot at the below shows the number of Gillespie algorithms(reactions) in 200 steps. The five stages of the viral growth process are simulated in each reaction. As the time length of each distinct stage is different,—generated from Exponential or Weibull distribution—there is no unique waiting time for each reaction. Therefore, it is not possible to provide the average for waiting times.

“Co-infection delay”(D) is the delay between the infection of the tissue by each strain. In other words, D is the number of steps at which the infection by the second strain starts. For example, D=30 means the second strain introduced to the model thirty steps after introduction of the first strain. Regards to the difference between the concepts of co-infection delay and waiting times, it is not possible to compare them.

C2-3) Figures and figure legends do not correspond to each other in some instances. I need explanation for Fig. 3 but I wonder where I can find it.

A2-3) We are sorry for the confusion we have caused. The figure numbers are now edited.

C2-4) In Fig. 6, B virus continues to accumulate but M virus does not. Could the authors provide intuitive explanation of this phenomenon? For example, limitation in the number of host cells will not explain because B virus would also suffer.

A2-4) The number of virions is not accumulated during the simulation. They are counted in each step, separately. To make more clarification, we have included some explanation about counting the virions and infected cells in the Method. The text reads:

“Infection calculation: Number of Virions and infected cells

The number of infected cells and free virions are calculated in each step of the simulation. The time length of each step is predefined before starting the simulation. Hence, the time length of the steps is constant. In our study, the time length of the simulation steps is about 1000 milliseconds. Co-Wish runs a number of Gillespie algorithms simultaneously. In other words, many cells can be infected at the same time.The time lengths of the algorithms with different settings are not alike. Each step is terminated when the sum of the time length of different runs of Gillespie algorithm is equal to the determined time length of the step.”

In our study, virus B—benign virus—has a shorter growth process than virus M. This property allows the virus to spread and reproduce faster than its competitor. Therefore, virus B can occupy more host cells. In other hands, when one cell is infected the cells in its vicinity are at risk to be infected by the same virus. Due to there is no assumption about host cell division, Virus B is a winner virus in the competition between two viruses and does not let virus M to infect the host cells. As it is obvious in Figure 6 in all assumed conditions, referred to different co-infection delay values for virus B, the slope of the plots are steeper for virus B.

C2-5) In relation to the above comment, I’m not sure if the authors assumed (1) limitation in the number of cells, (2) host cell division, and (3) innate immunity in each figure.

A2-5) More explanation about the simulated tissue has now been included in the “The dynamics of single strain infection” part of the revised version of the manuscript. The lines [138]-[141] of the revision now reads:

“To visualize the viral infection of tissue, Co-Wish simulates the tissue as a lattice in which each node represents one cell. The dimension of the lattice is determined by the user. In our simulations, the host cell division is not designed. In addition, there is no competition between host cells in Co-Wish. ”

The status of the immune system has now been included in the figure legends (Figures 2,4,6,7,8). In addition, the explanation about simulated immune system has now been revised:

“In Co-Wish, the immune system response is characterized by the probability at which the immune system can eliminate the toxic elements (γ), the delay between the infection and the immune response, and the capacity of the immune system (I). Toxic elements include one or both of the following: infected cells; virions. In other words, Co-Wish can eliminate the infected cells, the virions, as well as the combination of both. Here, the capacity refers to the maximum number of each kind of toxic element that can be eliminated by the immune system. The capacity value has to be less than the number of toxic elements that are potentially produced in the early steps of the simulation. γ is drawn from a uniform distribution, U(a; b), that its parameters are within [0; 1]. The immune system can eliminate toxic elements with a different probability. This characteristic indicates that all toxic elements produced in each step of the simulation do not have the same probability of elimination. For example, Co-Wish generates separate probability values for the killing of each virion. Here, we assume that a virion is killed with probability p and survives with probability 1 − p. ”

C2-6) In my understanding, authors think that competition for host cells is important in the simulation. Could the authors provide the time-course change of the number (and/or proportion) of cells attacked by both of the virus (before they get infected)? I’m not sure if the local spread of viruses assumed in this manuscript allows competition for host cells, though it totally depends on the virus and host cell densities assumed.

A2-6) The competition between two strains is over the host cells of the simulated tissue. The winner strain is the one which occupies more number of the cells than its competitor. In the proposed simulation model, the dimension of the lattice representing a tissue is determined before starting the simulation. The host division and competition between host cells is not assumed. Thus, the number of safe cells (but are potential to be infected by one of the two strains) is decreasing during the simulations. If one strain infects a cell, the cells in the vicinity of the infected cell are at risk to be infected by the strain. Thus, the rival strain has a poor possibility to infect those cells. That is how the competition between strains is constructed. The spread of infection follows a designed pattern. Inspired by the dispersion principle in physics, we designed the spread of the infection around one infected cell. The part “the spread of the viral infection” in Method describes the pattern. In addition, as it is mentioned in the (A2-5) we have now included more explanation about the simulated tissue to the revised version of the manuscript.

C2-7) L40 and L312: the authors touch incubation times of viruses but it was not clear to me how they are relevant to this study.

A2-7) The proposed simulation model does not consider the incubation times. Talking about the different incubation time of viruses in the introduction part is just to explain how “competitive exclusion principle” could be applied about co-infection of a tissue by different viral strains. To make more clarification about the simulation assumptions, we have now altered the explanation about the assumptions of our model in the introduction part. The introduction part in the revision, lines [11]-[41] now reads:

“The crux of this work is to couple the CEP with the life cycle of viruses. To model how two types of viruses, B and M, co-infect a tissue, we have to make a few assumptions:

1. Different viruses can have eclipse phases of varying lengths. The period between the initial infection and the first detectable viremia is known as the eclipse period—from the moment a virus enters the cell until it starts assembling its progenies and subsequences burst out of the cell. The duration of this period varies among different virus strains and affects the kinetic of infection (e.g., the eclipse period of SHIV lasts around a day [3] while it lasts between 7 to 8 hours in SARS-CoV [4].

2. Different viruses have different burst sizes, i.e., different per-cell virion particles. This assumption is reasonable, but even for SARS-CoV-2, the estimated burst size of 103 is simply based on MHV-2 data (e.g., [5]). The variation in particle-to-PFU among animal viruses [6] can be used to deduce the veracity of this assumption, but more direct data on SARS-CoV-2 is needed.

3. The CEP applies when two virus strains compete to infect a cell. In a spatially heterogeneous environment, different populations tend to partition the environment into non-overlapping micro-environments; for one of the most famous experimental demonstrations of such a situation see [9]. However, the displacement of one of the competitors by another is inescapable when the competitors cannot adapt or construct new niches in the environment in a reasonable timescale. The notion of CEP is undisputed when it comes to animals and bacteria trying to occupy the same ecological niche, but its application to viral confection is not as unequivocal as one would imagine.

In addition, different viruses can have different incubation times. For example, the median incubation period—i.e., the period between the onset of infection and the appearance of symptoms—for SARS-CoV-2 is estimated to be 5.1 days [7] (although such estimates should be taken with a grain of salt, e.g., [8]), whereas the incubation period for the common cold is around 1 − 3 days, and it could even be as long a few months to few years (e.g., Rabies and AIDS) [6]. Here, we do not consider any designation to simulate the different incubation times of two imaginary viruses.”

Other minor points:

C2-8) Figure 2 Legend: Replication -> replication

A2-8) It is now edited.

C2-9) L178: ith -> with

A2-9) It is now corrected.

C2-10) The reference formatting is terrible.

A2-10) It is now edited.

Point-by-Point responses to Reviewer #3

In this paper the authors have presented an interesting framework for the progression of viral infection in host tissue. The model simulates the invasion of two competing viruses in a theoretical lattice of initially susceptible cells. Cells that are infected by one virus are prevented by being simultaneously infected by a competing strain to simulate the effect of the competitive exclusion principle (CEP). The premise of this paper therefore may be appealing to not only mathematical modellers but also biologists and medics working in related fields.

However, I do have some key concerns, that if addressed would benefit the manuscript:

Thank you for the reviewer encouraging comments. We also thank you for your detailed review of our work and for your valuable and pertinent suggestions. We have edited the manuscript along the lines suggested by you. We believe that the presentation of our research has now been improved considerably. The responses to the comments are as follows.

C3-1) The authors do not provide a strong justification for the application of their model to SARS-CoV-2 in their introduction, which detracts from the narrative since the reader is left questioning the relevance of the model in this setting. Some justification is provided later in the discussion section, however, this is not particularly strong, and thus the application of the model to SARS-CoV-2 feels very extrapolative. I would therefore encourage the authors to consider removing the emphasis of the model’s application to SARS-CoV-2 in the abstract and introduction of the paper.

A3-1) The abstract and the introduction sections are now revised according to the reviewer ‘s suggestion. In the abstract of the revision text now reads:

“Inspired by the competition exclusion principle, this work aims at providing a computational framework to explore the theoretical feasibility of viral co-infection as a possible strategy to reduce the spread of a fatal strain in a population. We propose a stochastic-based model—called Co-Wish—to understand how competition between two viruses over a shared niche can affect the spread of each virus in infected tissue.

To demonstrate the co-infection of two viruses, we first simulate the characteristics of two virus growth processes separately. Then, we examine their interactions until one can dominate the other. We use Co-Wish to explore how the model varies as the parameters of each virus growth process change when two viruses infect the host simultaneously. We will also investigate the effect of the delayed initiation of each infection. Moreover, Co-Wish not only examines the co-infection at the cell level but also includes the innate immune response during viral infection.

The results highlight that the waiting times in the five stages of the viral infection of a cell in the model—namely attachment, penetration, eclipse, replication, and release—play an essential role in the competition between the two viruses. Finally, we discuss how co-infection with some forms of common respiratory diseases, such as common cold, might be useful in tackling a viral respiratory pandemic, such as the current COVID-19 crisis.”

At the beginning of the introduction:

“The “competitive exclusion principle” (CEP), also—perhaps erroneously—known as Gause’s law" [1], is the consequence of natural selection operating on non-interbreeding populations that occupy the same ecological niche. As Darwin put it, “ the competition will generally be most severe [...] between the forms which are most like each other in all respects” [2] (p.320). In the simplest reading, this principle implies that, in the competition between two sympatric non-interbreeding populations over the same ecological niche, one will displace the other.”

C3-2) The action of the CEP in the model was buried in the paper (it was briefly alluded to in lines 201 -203). This could have been made much clearer. Moreover, it appears that the authors are more specifically modelling superinfection exclusion, whereby secondary infection is prevented at a cellular level, which I would consider to be a specific example of the CEP.

A3-2) We agree with the reviewer that we should have put more emphasis on the CEP in our manuscripts. The introduction has been revised accordingly (as explained in A3-1). In particular, we highlighted the CEP and its application in modeling co-infection at a cellular level by addressing it at the beginning of the introduction.

C3-3) The paper would benefit if the model was described in clearer detail. For example, the authors have not made the action of the immune response in the model clear. I appreciate that to appeal to a wider audience that the authors may not wish to go into a great level of detail in the main body of the text, but I feel a more detailed description of all of the components of the model in the supplementary information would be very beneficial.

A3-3) The simulated immune system has been described in the “The influence of the immune system” part of the results section. In the revised manuscript, the explanations have now been edited and completed due to your comments (C3-10,C3-11,C3-12).

In addition, we have prepared a manual for the designed simulator. The manual file includes the description of all the components of the simulation model. It is now at:

https://github.com/safarvafadar/virusinfection

Parts of the manual explanation about immune system arguments are as follows:

Immune System Properties of the immune system

Description Activation of the immune system is optional. The designed immune system could resemble as innate response to the viral infection.

Arguments:

Killing probability A positive value from [0,1] to determine the probability at which immune system eliminates the virions or infected cells.

Delay A positive integer to specify the step at which the immune system activates.

Killing Mode The designed immune system is able to eliminate the infected cells (Whole-cell) or free virions (Part of the population) or a combination of both (Mix).

Capacity There are two options: Limited and Unlimited. If Limited is selected, a positive integer indicates the maximum number of infected items which will be eliminated by the immune system. The immune system will be deactivated after approaching the capacity. By selecting Unlimited, the immune system will be active to the end of the simulation.

C3-4) The paper would benefit from a more thorough justification of the parameterisation of the model, with suitable references to existing literature. While I appreciate different parameter combinations have been explored, this is somewhat limited, and I feel a more detailed sensitivity analysis would be beneficial. This also would most likely be appropriate to include in the supplementary information.

A3-4) Thanks for the suggestion. The results of simulations based on different values for the parameters plus more explanation about choosing the range of the parameters of the model have been now included in the supplementary file , part A. The plot at the below shows the results of different simulations based on ten values for the parameter λ for generating waiting times at replication stage:

To determine the range of the parameters value we ran different simulations, for a single virus model, by changing the parameter values and made different combinations of the different parameter values. Waiting times, the times that each virus spends in each stage of viral infection, were chosen from Exponential or Weibull distribution. To shorten the time length of one stage, we should increase the parameter λ of the Exponential distribution (λ>1) and decrease the shape parameter of the Weibull distribution(ω<1). After running different simulations, we figured out that shortening the time length of the viral infection process affects the number of infected cells (and produced virions) but the total trend is similar in all the simulations. Therefore, we decided to report just the simulation in which the parameter λ is 3 and the parameter ω is 0.5.

C3-5) The authors do not appear to use time units in their parametrisation of the model, or in the results presented, where all figures are presented in “steps” along the x-axis.

A3-5) The X-axis in all the figures represents the number of simulation steps (200 steps). We did not use the time units such as “second”. The complete explanation about the reason for using a number of simulation steps as timescale are in A3-6. Please, read the explanation at A3-6.

C3-6) The above point raises significant concerns – since the model is based on a Gillespie algorithm, each time step will be of a different time length due to different sojourn times between events. It could therefore be misleading to plot the trajectories of their results (e.g. number of virions) against the step count for each run of the simulation, since these are evolving over different timescales. I would therefore strongly advice that the model is revisited to ensure that all x-axes are presented over some units of time (hours or days).

A3-6) As you mentioned, the time length of each Gillespie algorithm to simulate the viral infection process is different. This characteristic of Gillespie algorithm is the main reason that we have displayed the results over the number of steps. In other words, in our simulations different Gillespie algorithms are run in each infected cell simultaneously. Since the time lengths of the different algorithms are not alike, it is difficult to calculate the total number of new infected cells (or virions) in each unit of time. To solve the problem, we decided to define the steps of the simulation with constant time length. In our study, the time length of the simulation steps is about 1000 milliseconds. One step is terminated when the sum of the time length of different Gillespie algorithms runs equal to the determined time length of the step. This designation gives us some suitable advantages:

- Providing the clearer view about the simulation process.

- Providing the possibility of comparing different simulations, since the results of all the simulations displayed over the same unit of time.)

- Providing the possibility of calculating the average number of infected cells and virions in each step.

The plots at the below show the time lengths and the number of Gillespie runs of 200 steps of one simulation. In the stepSum plot, there are slight differences between height of the bars because for some steps the sum of the time lengths of the Gillespie algorithms does not exactly equal to the determined time length for the step.

More explanations about the applied Gillespie algorithm have now been included in the manuscript. The introduction of the revision, lines [83]- [92] now reads:

“To simulate the stochastic multi-stage nature of the viral growth process occurring in viral co-infection, Gillespie algorithm, also known as the stochastic simulation algorithm, is usually considered as a trustworthy alternative. In particular, it supports simulating both discrete and stochastic behaviors of each stage of viral growth. The algorithm explicitly simulates the behavior of each virus in each stage of its growth process. Finally, it can provide a flexible framework to study the stochastic effects induced during the evolution of the infection. One characteristic of our simulation model is that several Gillespie algorithms run in infected cells, simultaneously. This characteristic can represent what we presume of viral infection in one tissue. Gillespie’s algorithms are indeed independent of their features.”

In the Method lines[438]-[446]:

“- Infection calculation: Number of Virions and infected cells

The number of infected cells and free virions are calculated in each step of the simulation. The time length of each step is predefined before starting the simulation. Hence, the time length of the steps is constant. In our study, the time length of the simulation steps is about 1000 milliseconds. Co-Wish runs a number of Gillespie algorithms simultaneously. In other words, many cells can be infected at the same time.The time lengths of the algorithms with different settings are not alike. Each step is terminated when the sum of the time length of different runs of Gillespie algorithm is equal to the determined time length of the step.”

C3-7) Additionally, a thorough spelling and grammar check is required.

A3-7) The manuscript has now been thoroughly edited.

While I do have a significant number of concerns with the manuscript in its current form, I do believe that a valuable piece of work could result from the model framework that the authors have used. Moreover, I felt the premise of the model is interesting in principle and would like to thank the authors for sharing their work.

I have made some additional inline comments below:

C3-8) line 69 - 70: If there is evidence of a reduced risk of a subsequent coronavirus infection by another respiratory virus then I think this needs to be explicitly stated to be able to back up the paper's justification for applying the model to SARS-CoV-2. Currently, it's not specified which pairs of viruses had a negative interaction between them in ref [10].

A3-8) More explanation about the viruses with negative interactions have been now included in the manuscript, introduction section lines [50]-[61]:

“There are two major lines of investigations that can illuminate the applicability of the CEP to viral infections in general, and SARS-CoV-2 in particular. These include (i) the experimental evidence on the negative interactions between viruses, and (ii) the computational models of co-infection. In relation to the first line of research, a recent paper [10] employed a population-level approach, based on 44230 cases over 9 years, to track the epidemiological interactions between 11 strains of respiratory viruses, including influenza A and B, rhinoviruses, and three human coronaviruses (229E, NL63, HKU1). The authors have shown both positive and negative interactions between pairs of viruses over time They inferred negative interaction between Rhinoviruses (A-C) and Influenza A virus—at both population and host levels—and negative interaction between Influenza A virus and Influenza B virus—at the population level. In addition, the three human coronaviruses showed negative interactions with Rhinoviruses (A-C), human Respiroviruses 1 and 4 at the population level.”

3-9) Line 133: The legend for figure 2 does not appear to match figure 2 in this document, which appears to show the different spreading patterns (square vs cross)

A3-9) Thank the reviewer for his careful consideration. It is now corrected.

C3-10) line 165: Does "the capacity of the immune system (I)" have any units, such as the maximum number of infected cells/virions eliminated per day? It is not clear how "I" is defined mathematically in the model.

A3-10) The parameter “I” is the maximum number of virions (or infected cells) that would be eliminated by the immune system in each simulation step. It is a positive integer and its value is chosen empirically. Determining the capacity depends on the number of virions (or infected cells) which are produced in the early steps of the simulations, the dimensionality of the lattice. The capacity value should be chosen less than the number of virions which potentially can be produced at early steps of the simulation. Thus, in each step some virions would be allowed to be survived. In our simulations, “I” is set as 100 because during different simulations we figured out in each step the number of produced virions is more than 100.

It is also worth mentioning that in the Co-Wish simulator user is able to set different values for parameter I. More explanation about setting the simulation parameters such as “I” are included in the prepared manual of the simulator.

In addition, more explanation about the parameter “I” has been included in the manuscript line[158] – [166]. The text now reads:

“ In Co-Wish, the immune system response is characterized by the probability at which the immune system can eliminate the toxic elements (γ), the delay between the infection and the immune response, and the capacity of the immune system (I). Toxic elements include one or both of the following: infected cells; virions. In other words, Co-Wish can eliminate the infected cells, the virions, as well as the combination of both. Here, the capacity refers to the maximum number of each kind of toxic element that can be eliminated by the immune system. The capacity value has to be less than the number of toxic elements that are potentially produced in the early steps of the simulation. ”

C3-11) line 167: Gamma appears to have been defined in lines 163-164 as "the rate at which the immune (system) eliminates infected cells or virions". Therefore, it presumably has units of "cells/virions cleared per hour/day". I therefore do not understand the logic behind why it is then drawn from a U [0,1] distribution.

A3-11) The simulated immune system eliminates each virion (or infected cell) with different probability. In other words, all the virions produced in each step of the simulation do not have the same probability to be eliminated. The simulator generates a separate probability value for each virion (or infected cell). The value “ p ” for the parameter γ means the virion will be eliminated with probability “ p ” and it will be survived with probability “1-p”.

In the revised manuscript, we have replaced the word ‘rate’ with the word ‘probability’. In addition, we have now included more explanation about the parameter γ to the ‘The influence of the immune system’ part of the manuscript:

“In Co-Wish, the immune system response is characterized by the probability at which the immune system can eliminate the toxic elements (γ), the delay between the infection and the immune response, and the capacity of the immune system (I). Toxic elements include one or both of the following: infected cells; virions. In other words, Co-Wish can eliminate the infected cells, the virions, as well as the combination of both. Here, the capacity refers to the maximum number of each kind of toxic element that can be eliminated by the immune system. The capacity value has to be less than the number of toxic elements that are potentially produced in the early steps of the simulation. γ is drawn from a uniform distribution, U(a; b), that its parameters are within [0; 1]. The immune system can eliminate toxic elements with a different probability. This characteristic indicates that all toxic elements produced in each step of the simulation do not have the same probability of elimination. For example, Co-Wish generates separate probability values for the killing of each virion. Here, we assume that a virion is killed with probability p and survives with probability 1 − p. ”

C3-12) Line 176: to differentiate this from the "unlimited immune response capacity" situation, I think it would be better to have notation along the lines of 0.

A3-12) We have modified the notations. In the part of “The influence of the immune system” of the revision text now reads:

“ 1. Weak with limited capacity (WL): γ ∼ U(0; 0:4)

2. Weak with unlimited capacity (WU): γ ∼ U (0; 0:4)

3. Strong with limited capacity (SL) : γ ∼ U(0:6; 1)

4. Strong with unlimited capacity (SU) : γ ∼ U(0:6; 1)

Limited capacity indicates that the immune system will not be active in all the steps of the simulation. It will be deactivated when the capacity hits the predetermined value for I. If the capacity is unlimited, the immune system will be active by the end of simulation.”

C3-13) line 199: It is unclear how the second virus is seeded into the model. It would be helpful if the authors clarified this (e.g. is it seeded into a cell in the lattice at random?)

A3-13) The location of the second virus is chosen randomly. We have now included more explanation about locating the second virus to the “Competition during co-infection” part. The text now reads:

“The location of the second virus in the lattice is chosen randomly, and it is independent of the first virus location. Those cells that are not infected by the first virus can be a niche of the second virus.”

C3-14) line 213: There is a lack of justification for the parameters chosen for the various distributions in table 1. While there may be limited literature for SARS-CoV-2, literature on other respiratory viruses may allow the authors to justify their choices. As highlighted before, I feel the paper would benefit if all times are given in appropriate units such as days, rather than being treated as dimensionless. I would also encourage the authors to consider including a sensitivity analysis of the different parameters in the supplementary information.

A3-14) As we have mentioned in (A3-4) more explanation about choosing the parameters value and the results of more simulations with different values of the parameters have been now included in the supplementary files. To design the viral infection process, we reviewed different literature about different viruses. Reviewing the studies informed us about the limitations of deterministic mathematical methods such as ODE. Therefore, we decided to propose a stochastic approach to study the viral (co-)infection. A comparison between deterministic and stochastic models are in the introduction part,lines[75]-[89].

The arguments of the model were determined based on the different studies about viruses including SARS-CoV-2 and other respiratory viruses. For example, we figured out that the eclipse stage, among different stages of viral infection, has a strong influence on the process. Therefore, to generate the waiting times of the eclipse phase we applied Weibull distribution considering the shape of its hazard function. Using Weibull distribution gives us the ability to control the hazard rate of the event of transition to the next stage. The explanation about the reason for applying Weibull distribution is in the main text lines [368] – [384].

As it is explained in the (A3-6) we displayed the results of the simulations over the number of steps. We hope our explanations in A3-6 could clarify the reasons for using a number of steps as timescale.

C3-15) lines 267-271: Hepatitis G is now typically referred to as GB virus C.

A3-15) It is now corrected.

C3-16) lines 312-314: If the authors wish to keep the focus of the paper on SARS-CoV-2, I feel the manuscript would benefit if these studies (refs [30] and [31]) were highlighted in the introduction section. The reader would then understand the authors justification for the model to be placed in the context of SARS-CoV-2.

A3-16) Following the reviewer’s suggestions in C3-1 and C3-2, the manuscript, the abstract, and introduction parts have now been revised. We believe that the presentation of our research has now been improved considerably.

Attachment

Submitted filename: Response to Reviewers.pdf

Decision Letter 1

Bashar Ibrahim

9 Dec 2020

PONE-D-20-22300R1

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: a computational perspective

PLOS ONE

Dear Dr. Sadeghi,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jan 23 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

  • A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

  • An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Bashar Ibrahim

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #3: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #3: N/A

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The corrections regarding my revision were done well. I would agree to the publication of the paper now.

Reviewer #3: I would like to thank the reviewers for addressing the comments myself and the other reviewers had made in response to the manuscript's initial submission, and believe the presentation of the work has improved as a result. I do nevertheless have some ongoing concerns:

1) While I am thankful that the authors have given an improved explanation behind their choice of parameters in the supplementary information, I still believe that the paper would benefit significantly if these were placed in the context of some real world viruses. Currently the choice of parameters, both in the main body of the text and the supplementary information, are entirely theoretical and have not been justified. I feel this would be acceptable if the paper was presented as a wholly theoretical piece of work, however, the authors ultimately conclude that "Co-Wish suggests that common respiratory viral infections - such as common human coronaviruses - may limit the replication of SARS-CoV-2" (lines 317-318). Yet, when the choice of parameters is entirely theoretical, I do not believe such conclusions should be drawn, or even alluded to.

2) The results of the paper are still presented over steps rather than units of time. While I appreciate the authors responded to this point when I raised it previously, I still feel that it would be important to translate the results to be presented over time units if conclusions such as the one highlighted above are to be drawn. Moreover, it should not be too difficult for the authors to record the sojourn times between events from a Gillespie algorithm.

While I still have these concerns, I do hope that the authors are not too disheartened and want to stress that I still believe it is an interesting piece of work. I do however believe that the authors should consider taking one of either of the following approaches:

- Revisiting their model to present their results in units of time, with a choice of parameters that have been justified in some way from common respiratory viruses. I appreciate that there will be a lot of unknowns for the waiting times between different stages for real world viruses, but I believe it would be possible to make at least sensible ballpark estimates. If this were to be done, then I feel the authors would have more justification to make conclusions that could be placed in a real world context.

- However, if the authors do not wish to re-run their model and results, as suggested above, I feel the narrative of the paper should be revisited to emphasize that it is an entirely theoretical piece of work, and that any conclusions in relation to any real world viruses should not be made or alluded to.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #3: Yes: Andrew Glover

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2021 Feb 24;16(2):e0247200. doi: 10.1371/journal.pone.0247200.r004

Author response to Decision Letter 1


9 Jan 2021

Reviewer #3: I would like to thank the reviewers for addressing the comments myself and the other reviewers had made in response to the manuscript's initial submission, and believe the presentation of the work has improved as a result. I do nevertheless have some ongoing concerns:

We would like to thank the reviewer’s suggestions and encouraging words. The responses to his comments are as follows:

C1) While we are thankful that the authors have given an improved explanation behind their choice of parameters in the supplementary information, we still believe that the paper would benefit significantly if these were placed in the context of some real world viruses. Currently the choice of parameters, both in the main body of the text and the supplementary information, are entirely theoretical and have not been justified. I feel this would be acceptable if the paper was presented as a wholly theoretical piece of work, however, the authors ultimately conclude that "Co-Wish suggests that common respiratory viral infections - such as common human coronaviruses - may limit the replication of SARS-CoV-2" (lines 317-318). Yet, when the choice of parameters is entirely theoretical, I do not believe such conclusions should be drawn, or even alluded to.

A1)

We followed your suggestion; accordingly, we have edited the manuscript to emphasize the theoretical aspects of the study.

In the Abstract of the revision, the text reads:

“While it could prove challenging to fully understand the therapeutic potentials of viral co-infection, we discuss that our theoretical framework hints at an intriguing research direction in applying co-infection dynamics in controlling any viral outbreak’s speed.”

In the Introduction of the revision, the paragraph below is now omitted:

“Based on the above assumptions, we propose that the competitive exclusion principle can, at least theoretically, be useful in tackling a viral pandemic, such as the current COVID-19 crisis, since one viral strain might be able to out-compete and eliminate a rival strain in the same ecological niche. Considering the current pandemic, this question offers hope as a means to control the novel coronavirus outbreak by supplanting the SARS-CoV-2 virus by a much less fatal one.”

The Discussion section has been edited as well; for example, some explanations about the SARS-CoV-2 virus is now deleted.

● At the beginning of the discussion, the following text is now dropped:

“We looked into the dynamics of viral co-infection to investigate the possibility of eliminating malignant virus strains in the population by pitting it against a more benign strain.”

The text lines [230]-[233] now reads:

“In this study, we look into the dynamics of viral co-infection using a CEP-inspired computational model. To explore the possible effects of co-infection by two virus species, we develop a stochastic-based viral infection simulator—called Co-Wish—to model and display the competition between two strains.”

● In the discussion, the paragraph below is now omitted:

“Following the same line of thought, we examined the possibility of slowing down the spread of COVID-19 through the introduction of a more harmless virus to a population. Our results offer the possible mitigating role of benign respiratory virus infection as a way of antiviral treatment in COVID-19 individuals. Accordingly, there is a clear and pressing need to conduct clinical trials to assess the role of co-infection in the clinical outcomes of SARS-CoV-2 infected patients.”

● Revised discussion focuses more on the theoretical aspects of the proposed models. We only discuss that our theoretical framework hints at an intriguing research direction in the dynamics of co-infection. Accordingly, the fifth and sixth paragraphs in the previous version are summarized into one paragraph—highlighted in green in the revised manuscript—which only concisely and suggestively discusses the possibility of controlling the spread of any viral infection, such as the SARS-CoV-2 virus, in the population by supplanting the lethal virus with a much less fatal one.

C2) The results of the paper are still presented over steps rather than units of time. While I appreciate the authors responded to this point when I raised it previously, I still feel that it would be important to translate the results to be presented over time units if conclusions such as the one highlighted above are to be drawn. Moreover, it should not be too difficult for the authors to record the sojourn times between events from a Gillespie algorithm.

A2)

Regarding the definition of one step in our simulation, we present the results over almost equal-length segments of assumed time. One step is terminated when the sum of the time length of different runs of the Gillespie algorithm equal to, at least, the determined time length of steps. In our simulation, the length of steps is about 1000msec; we represent the results once every 1000msec.

The summary of our reasons for displaying the results by using step number are as follows:

- Avoiding Computational Costs:

Using Discrete-Time Simulation (DTS), for example, displaying the results at each millisecond, imposes a lot of computational overhead such as CPU,IO, and RAM overhead. The computational costs make the simulation slow.

Instead, we apply the TIME ADVANCE MECHANISM (TAM) approach and a parallel timeline for each infected cell viral growth process.

- Preparing Concise view about the parallel simulations:

We believe that presenting the results over a number of steps can illustrate a more concise view about the parallel simulations held in different virtual infected cells than showing them over time. The main reason for this decision is simulations at infected cells are in parallel with different lengths of time. Showing the results over the number of steps allows the viral growth process to get finished in more infected cells. Figure 1 displays the main idea of the simulation.

Figure 2 shows the simulation results over time unit (millisecond) just for one run. In our study, each simulation includes one hundred runs. The trajectory is very dense because it represents a large number of points. It seems it does not convey more information than displaying the results over the number of steps.

Figure 1- schematic picture of viral infection simulation. Simulations in “n” cells are simultaneous. Different clocks illustrate different time lengths of the simulations. “Evaluate Time to Infect Neighbors” checks out the time length of the simulation.

Figure 2- simulation results over time unit (millisecond) just for one run

Attachment

Submitted filename: Response to Reviewer #3 comments.docx

Decision Letter 2

Bashar Ibrahim

13 Jan 2021

PONE-D-20-22300R2

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: a computational perspective

PLOS ONE

Dear Dr. Sadeghi,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a minor revised version of the manuscript that addresses the points raised during the review process.

"Please elaborate more about the difference between your approach and other literature theoretical approaches particularly there is a huge number recent models and also software tools for SARS-CoV-2 and some on a related study. This would help and guide readers more in the introduction or discussion. I listed a few examples and more is even better in this case:   

10.20944/preprints202005.0376.v1 

10.1155/2020/4923856

10.1007/s00018-019-03382-0

https://doi.org/10.1016/j.bihy.2008.05.003

10.3390/v13010014

"

Please submit your revised manuscript by Feb 27 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

  • A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

  • An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Bashar Ibrahim

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2021 Feb 24;16(2):e0247200. doi: 10.1371/journal.pone.0247200.r006

Author response to Decision Letter 2


31 Jan 2021

Dear Professor Bashar Ibrahim

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: a computational perspective

We are thankful for your consideration of our manuscript and for the constructive point you mentioned. We have attached the revised version of our manuscript (with modifications tracked in blue).

This version of the manuscript has followed the suggestion we have received from the editor. Accordingly, we have discussed and analyzed the proposed model with other theoretical approaches. The third version of our manuscript now includes two dedicated paragraphs explaining the differences between the proposed method and other available theoretical approaches. These new paragraphs have been incorporated into the discussion section of the revised version.

We hope that the revised manuscript could address the comment pointed out by you.

We would like to thank you and the reviewers again for the careful reading of our work and the excellent comments we have received.

Mehdi Sadeghi

Associate Prof.

National institute of genetic engineering and biotechnology

Tehran

Iran

Decision Letter 3

Bashar Ibrahim

3 Feb 2021

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: a computational perspective

PONE-D-20-22300R3

Dear Dr. Sadeghi,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Bashar Ibrahim

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Acceptance letter

Bashar Ibrahim

11 Feb 2021

PONE-D-20-22300R3

Competitive exclusion during co-infection as a strategy to prevent the spread of a virus: a computational perspective

Dear Dr. Sadeghi:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Bashar Ibrahim

Academic Editor

PLOS ONE

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    S1 File

    (PDF)

    Attachment

    Submitted filename: Response to Reviewers.pdf

    Attachment

    Submitted filename: Response to Reviewer #3 comments.docx

    Data Availability Statement

    The program written simulate the spread of the competition between viruses is available at https://github.com/safarvafadar/virusinfection.


    Articles from PLoS ONE are provided here courtesy of PLOS

    RESOURCES