Nonlinear Dynamics of Sensitive-Resistant Tumor Populations under Therapy-Induced Clonal Competition
Panyawat Haarsa* Author for corresponding; e-mail address: chaiwichith@g.swu.ac.th
ORCID ID: https://orcid.org/0000-0003-0622-4411
Volume: Vol.53 No.6 (November 2026: In progress)
Research Article
DOI: https://doi.org/10.12982/CMJS.2026.098
Received: 14 June 2026, Revised: 27 August 2026, Accepted: 14 September 2026, Published: 27 October 2026
Citation: Haarsa P., Nonlinear dynamics of sensitive-resistant tumor populations under therapy-induced clonal competition. Chiang Mai Journal of Science, Year; 53(6): e2026098. DOI 10.12982/CMJS.2026.098.
Graphical Abstract
Abstract
This study develops and analyzes a nonlinear mathematical model for therapy-sensitive and therapy-resistant tumor populations under therapy-induced clonal competition. The model incorporates logistic growth, asymmetric interclonal competition, conversion from sensitivity to resistance, and a constant treatment pressure acting directly on sensitive cells. We establish well-posedness of the model by proving positivity, boundedness, global existence, and the existence of an absorbing region in the biologically feasible domain. Equilibrium points and threshold quantities are derived to characterize sensitive-cell persistence, resistant dominance, coexistence, and resistant-cell invasion in the mutation-free case. Local stability and bifurcation-related conditions are analyzed using the Jacobian matrix and threshold functions. The results show that increasing treatment intensity can suppress sensitive cells while reducing their competitive pressure on resistant cells. This mechanism promotes resistant dominance through therapy-induced competitive release. Numerical simulations support the analytical findings. Weak treatment may lead to coexistence, whereas strong treatment can produce an initial decline in tumor burden followed by resistant regrowth. Bifurcation-style and heat-map simulations further show how treatment intensity and clonal competition shape long-term tumor composition. These findings provide a threshold-based framework for understanding therapy-induced resistance and adaptive treatment design.
1. INTRODUCTION
Cancer is not only a disease of uncontrolled growth. It is also an evolutionary process. A tumor contains diverse cell populations. These populations compete for space, nutrients, and other local resources. They may also differ in their sensitivity to treatment. This heterogeneity is one reason why long-term cancer control remains difficult. Drug-sensitive cells may decline under therapy. Resistant cells may then survive and expand. This process can lead to relapse and treatment failure.
The evolutionary view of cancer has a long history. Nowell [1] proposed that tumor progression is driven by clonal evolution. This idea has shaped modern cancer biology. Later studies emphasized that cancer progression can be interpreted through both evolutionary and ecological perspectives [2]. Tumor microenvironmental constraints can also shape ecological interactions among cancer cell populations [3]. Broader classifications of neoplasms further emphasize evolutionary and ecological features of cancer [4]. Within this view, cancer progression results from mutation, selection, competition, and adaptation. Later studies further highlighted that therapeutic intervention can impose strong selective pressure [5–7]. This pressure may favor resistant variants. Hence, treatment can reduce tumor burden in the short term but increase the relative advantage of resistant clones in the long term.
Mathematical models provide a useful framework for studying tumor growth, treatment response, resistance, and therapy optimization [8–11]. They help isolate mechanisms that are difficult to observe directly and allow comparison of treatment strategies. However, many basic models treat the tumor as a single homogeneous population. Such models describe overall growth but are less suitable for studying competition between sensitive and resistant cells. A more informative approach is to represent the tumor as interacting subpopulations. In this framework, sensitive and resistant cells have distinct growth rates and compete for shared resources. This structure is consistent with Lotka-Volterra competition models and reflects the ecological nature of tumor growth. When treatment is weak, sensitive cells may suppress resistant cells through competition. Under strong therapy, sensitive cells may decline. This can reduce competitive pressure and allow resistant cells to expand, leading to competitive release.
This mechanism is central to adaptive therapy. Gatenby et al. [12] proposed adaptive therapy as an alternative to maximum tolerated dose treatment. Subsequent studies further developed this idea from evolutionary, game-theoretic, and treatment-timing perspectives [13–15]. The aim is not always to eliminate all sensitive cells. Instead, sensitive cells may be preserved to suppress resistant cells through competition. This idea has motivated a large body of theoretical and experimental work. Spatial models have shown that the balance between sensitive and resistant populations can strongly influence treatment outcomes [16–18]. Recent mathematical work has also compared intermittent and continuous adaptive therapy using control-theoretic arguments [19]. These studies show that competition structure is not a technical detail. It can determine whether a treatment strategy delays or accelerates resistance.
Despite these advances, several modeling gaps remain. Many studies focus on treatment scheduling but give less attention to the qualitative dynamics of sensitive-resistant competition. The roles of interclonal competition coefficients are also not always classified. These coefficients determine whether one clone excludes the other or whether coexistence is possible. In addition, therapy-induced selection is often discussed biologically but is not always connected to explicit threshold conditions. Such thresholds are useful because they identify parameter regimes where treatment may suppress tumor burden while promoting resistant dominance. Rigorous analysis of positivity, boundedness, equilibria, and stability therefore remains important.
The present study addresses these gaps by analyzing a nonlinear two-population model of therapy-sensitive and therapy-resistant tumor cells. The model includes logistic growth, asymmetric clonal competition, mutation from sensitivity to resistance, and treatment acting on sensitive cells. In this baseline model, the treatment intensity is treated as a fixed parameter. It is not yet optimized as a time-dependent decision variable. This choice allows the core dynamical mechanisms to be examined clearly before optimal-control extensions are introduced. The main contributions are fourfold. First, we formulate a tractable sensitive-resistant competition model. Second, we establish positivity, boundedness, global existence, and an absorbing region in a biologically meaningful domain. Third, we derive equilibria, threshold quantities, and local stability conditions. Fourth, we use numerical simulations to illustrate therapy-induced competitive release and long-term tumor composition.
The novelty of this study lies in its threshold-based analysis of therapy-induced clonal competition. The model is studied within a fixed-treatment framework. The treatment intensity is not optimized in this baseline model. The model links treatment pressure, mutation, and asymmetric interclonal competition through explicit threshold quantities. These thresholds describe sensitive-cell persistence, resistant dominance, coexistence, and resistant-cell invasion in the mutation-free case. They also indicate when treatment may suppress sensitive cells but promote resistant expansion through competitive release. The study combines well-posedness results, equilibrium analysis, stability results, bifurcation thresholds, and numerical simulations. It therefore provides a compact framework for interpreting therapy-induced resistance. It also motivates future extensions to adaptive therapy, optimal control, and spatial modelling.
2. MODEL FORMULATION
We formulate a deterministic model for two interacting tumor cell populations. The first population is sensitive to therapy. The second population is resistant to therapy. The model is intended as a minimal framework. It is not designed for a specific cancer type. It is designed to study clonal competition under treatment pressure.
The formulation is motivated by three biological facts. First, tumor growth is limited by local resources. These resources include space, oxygen, nutrients, and access to blood supply. Logistic and Gompertz-type equations are often used to describe tumor growth under resource limitation [20–21]. Second, tumors are heterogeneous. They contain subclones with different fitness values and different responses to therapy [5]. Third, treatment may suppress sensitive cells while allowing resistant cells to expand. This idea is central to mathematical studies of resistance and adaptive therapy [12, 22].
Let \(S(t)\) denote the density of therapy-sensitive tumor cells at time \(t\). Let \(R(t)\) denote the density of therapy-resistant tumor cells. Both variables are non-negative. The total tumor burden is \(S(t)+R(t)\). The model assumes that both populations grow logistically. However, the strength of competition is allowed to differ between the two populations. This is represented by the competition coefficients \(\alpha\) and \(\beta\). The coefficient \(\alpha\) measures the effect of resistant cells on sensitive cells. The coefficient \(\beta\) measures the effect of sensitive cells on resistant cells.
The proposed model is given by
\[ \frac{dS}{dt}=r_sS\left[1-\frac{S+\alpha R}{K}\right]-\mu S-uS, \tag{2.1} \]
\[ \frac{dR}{dt}=r_rR\left[1-\frac{R+\beta S}{K}\right]+\mu S. \tag{2.2} \]
Here, \(r_s\) and \(r_r\) are the intrinsic growth rates of the sensitive and resistant populations. The parameter \(K\) is the environmental carrying capacity. The parameter \(\mu\) is the transition rate from the sensitive state to the resistant state. The parameter \(u\) is the treatment intensity. In this baseline model, \(u\) is a fixed non-negative treatment parameter. It is not optimized as a time-dependent decision variable. This choice keeps the model analytically tractable. It also provides a baseline for future optimal-control extensions.
The first equation describes the dynamics of sensitive cells. The growth term represents proliferation under resource limitation. It also includes competition from resistant cells through the coefficient \(\alpha\). The loss term \(-\mu S\) represents conversion from sensitivity to resistance. The loss term \(-uS\) represents therapy-induced killing of sensitive cells. Thus, the sensitive-cell equation includes proliferation, resource limitation, conversion, and treatment-induced suppression.
The second equation describes the dynamics of resistant cells. Its growth term represents proliferation under resource limitation and competition from sensitive cells through the coefficient \(\beta\). The term \(+\mu S\) represents the inflow of resistant cells from the sensitive population.
The model assumes that resistant cells are not killed directly by therapy. This is a simplifying assumption. It represents the limiting case of complete resistance. It also allows a clear analysis of therapy-induced competitive release. In this setting, treatment reduces the sensitive population. This may reduce competition for space and resources. As a result, resistant cells may gain a relative ecological advantage even when their intrinsic growth rate is lower. In clinical terms, competitive release means that therapy may reduce the sensitive population and thereby free ecological resources for resistant cells. This phenomenon has been discussed in evolutionary and adaptive therapy approaches to cancer treatment [5, 12, 16–18].
2.1 Variables and Parameters
The variables and parameters used in the model are summarized in Tables 1 and 2.
2.2 Modelling Assumptions
The model is based on simplifying assumptions. These assumptions keep the system analytically tractable. They also preserve the main biological mechanisms.
The tumor is represented by two dominant phenotypic groups. These are therapy-sensitive cells and therapy-resistant cells. Minor subclones are not modelled explicitly. They are absorbed into these two effective classes.
Both populations grow in the same tumor microenvironment. They share a common carrying capacity. Thus, they compete for space, nutrients, oxygen, and other local resources. The two populations may have different intrinsic growth rates. This allows the model to represent a possible fitness cost of resistance.
Competition is asymmetric. The coefficient measures the effect of resistant cells on sensitive cells. The coefficient measures the effect of sensitive cells on resistant cells. These coefficients need not be equal.
Therapy acts directly on sensitive cells. Resistant cells are not killed directly in this baseline model. Resistance is acquired in one direction, from sensitive cells to resistant cells. Back conversion is neglected. These assumptions yield a nonlinear system that can describe coexistence, resistant dominance, threshold behavior, and therapy-induced competitive release.
2.3 Dimensionless Form
For analysis, it is useful to reduce the number of parameters. We introduce the dimensionless variables and. We also rescale time by The following dimensionless parameters are used:
The dimensionless model becomes
(2.3)
(2.4)
This form is useful for stability and bifurcation analysis. The parameters , and define treatment pressure, conversion, resistant fitness, and competition. Together, they describe how therapy and clonal competition shape long-term tumor composition.
2.4 Initial Conditions and Feasible Region
The model is studied with non-negative initial conditions,
The biologically feasible region is the non-negative quadrant. Solutions should remain non-negative because negative cell densities have no biological meaning. In this baseline model, is constant. Later extensions may replace with a time-dependent control . They may also include treatment-dependent conversion and spatial diffusion. These extensions are not part of the present baseline analysis.
3. WELL-POSEDNESS AND BIOLOGICAL FEASIBILITY
In this section, we establish the basic dynamical properties of the proposed model. These results are required before the equilibrium and stability analysis can be performed. Since and represent tumor cell populations, the solutions must remain non-negative. They must also remain finite for all future time. These properties ensure that the model is biologically meaningful and mathematically well-posed.
For the analysis in this section, we use the dimensional model introduced in Section 2. It is restated here only to make the proofs self-contained:
(3.1)
All variables and parameters are defined in Section 2. We assume throughout this section that all parameters are non-negative, with , and The treatment intensity is fixed and non-negative. The results in this section follow standard arguments from ordinary differential equations and population dynamics. These arguments are commonly used to verify biological feasibility in mathematical biology models [21, 23–25].
3. WELL-POSEDNESS AND BIOLOGICAL FEASIBILITY
3.1 Local Existence and Model Consistency
We first show that the model has a unique local solution for any biologically admissible initial condition.
Theorem 3.1 For any initial condition system (3.1) has a unique local solution. In particular, this holds for all biologically admissible initial conditions in
Proof. The right-hand side of system (3.1) is a polynomial vector field in and Hence, it is continuously differentiable and locally Lipschitz on The standard existence and uniqueness theorem for ordinary differential equations then gives a unique local solution for any initial condition in ℝ2. This proves the result.
This result ensures that each initial tumor composition generates a well-defined trajectory. Biologically, a fixed initial mixture of sensitive and resistant cells gives a unique short-term prediction under fixed parameter values and a constant treatment intensity
3.2 Positivity and Biological Feasibility
We now show that the non-negative quadrant is positively invariant. This means that solutions starting with non-negative initial data remain non-negative for all future time.
Define
Theorem 3.2 Assume that and Then the solution of system (3.1) satisfies and for all for which the solution exists.
Proof. We examine the vector field on the boundary of On the boundary the first equation gives Hence, the solution cannot cross from into . On the boundary the second equation gives since and Thus, the vector field is tangent to or points into Therefore, the non-negative quadrant is positively invariant.
This result ensures that the model never produces negative tumor cell populations. Thus, therapy and mutation keep the system inside the biologically feasible region.
3.3 Boundary Behavior and Emergence of Resistance
The boundary dynamics provide useful biological information. In particular, the two coordinate axes do not have the same dynamical meaning. The sensitive-free boundary is invariant. The resistant-free boundary is generally not invariant when mutation is present.
Proposition 3.1 The boundary is invariant for system (3.1). The boundary is invariant when If then any point on with leaves the resistant-free boundary and enters the interior of the non-negative quadrant.
Proof. On , the first equation gives . Hence, trajectories starting on remain on this boundary. On the second equation gives If then on so the boundary is invariant. If and then . Thus, trajectories leave the resistant-free boundary and enter the interior of the non-negative quadrant. This proves the result.
This proposition shows that resistance can emerge from a purely sensitive tumor population when mutation is present. Thus, the complete absence of resistant cells is not dynamically robust.
3.4 Uniform Boundedness and Resource Limitation
We next prove that the total tumor population remains bounded. This result rules out unbounded growth and ensures that the long-term dynamics occur in a finite region.
Let
denote the total tumor cell population. Adding the two equations of system (3.1), we obtain
(3.2)
The mutation terms cancel in the equation for . This is biologically reasonable. Mutation changes the phenotype of tumor cells. It does not directly create new tumor mass in the present formulation.
Theorem 3.3 All solutions of system (3.1) with non-negative initial conditions are ultimately bounded. More precisely, if then
where
Proof. Let and Since the mixed competition terms are non-positive, we obtain
Using It follows that
(3.3)
This logistic-type differential inequality implies, by comparison, that
Since the result follows.
This result shows that the total tumor burden cannot grow without bound. Resource limitation and competition restrict the population. Therapy may change tumor composition, but the total population remains bounded.
3.5 Global Existence
The local solution obtained in Theorem 3.1 can now be extended to all future time. This follows from positivity and boundedness.
Corollary 3.1. For every initial condition system (3.1) has a unique global solution. That is, the solution exists for all .
Proof. By Theorem 3.1, system (3.1) has a unique local solution. By Theorem 3.2, the solution remains in the non-negative quadrant for as long as it exists. Hence,
Let
From equation (3.2), and using and we have Let Then Therefore, By Gronwall’s inequality, for every finite
Thus, and remain finite on every finite time interval. Hence, the solution cannot blow up in finite time and extends globally for all . Biologically, the model can be used to study long-term tumor evolution and resistance emergence. This result also supports the later stability analysis.
3.6 Compact Absorbing Region and Long-Term Tumor Dynamics
We finally show that the system has a compact absorbing set in the non-negative quadrant. This result is useful for the later analysis of equilibria and long-term dynamics.
Let
Choose any constant such that Define
Theorem 3.4 The set is a compact absorbing set for system (3.1). More precisely, for every solution with initial condition in there exists a time such that for all .
Proof. By Theorem 3.3, Since the definition of the limit superior implies that there exists a time such that for all By Theorem 3.2, for all . Therefore,
for all Hence, absorbs every solution starting in the non-negative quadrant. The set is closed and bounded in Therefore, it is compact. This proves the theorem.
Theorem 3.4 shows that all trajectories eventually enter and remain in the absorbing region Thus, the total tumor burden stays below a finite upper level. Biologically, represents a long-term admissible tumor region constrained by limited ecological resources. This supports later equilibrium, stability, and bifurcation analyses.
3.7 Summary of The Basic Properties
The results in this section show that system (3.1) is well-posed in the non-negative quadrant. For every non-negative initial condition, the model has a unique global solution. Both tumor cell populations remain non-negative, and the total tumor burden is uniformly bounded. All trajectories eventually enter a compact absorbing set.
These properties justify the subsequent dynamical analysis. They also show that the model is consistent with an ecological view of tumor growth. Sensitive and resistant cells compete within a finite microenvironment. Treatment suppresses sensitive cells. Mutation transfers cells to the resistant class. Such baseline models also help connect mathematical theory with clinically oriented open questions in adaptive therapy [26].
4. EQUILIBRIA AND THRESHOLD QUANTITIES
In this section, we identify the equilibrium points of the model. The model was formulated in Section 2 and restated in Section 3 for the well-posedness analysis. We therefore do not repeat the full system here. Instead, we use the equilibrium conditions derived from system (3.1). These conditions characterize possible long-term tumor compositions. An equilibrium point satisfies
(4.1)
Thus
(4.2)
and
(4.3)
The first equation shows that either or the sensitive population satisfies a balance between growth, competition, mutation, and treatment. The second equation shows that the resistant population can be maintained by its own growth and by the inflow from sensitive cells through the mutation term
4.1 Tumor-free Equilibrium
The tumor-free equilibrium is This equilibrium always exists. It represents the complete absence of both sensitive and resistant tumor cells. The tumor-free state is mathematically valid. However, it should be interpreted with caution in this model. The therapy term acts only on sensitive cells. There is no direct killing term for resistant cells. Hence, the tumor-free state is not expected to be a robust long-term outcome when resistant cells have positive growth potential. This reflects a limiting case in which treatment does not eliminate resistant cells directly.
4.2 Resistant-only Equilibrium
Setting in equation (4.3) gives
Thus, either which gives , or Hence, the resistant-only equilibrium is This equilibrium always exists. The equilibrium represents a tumor population fully dominated by resistant cells. At this state, the sensitive population has disappeared. The resistant population remains at the carrying capacity of the tumor microenvironment. This equilibrium is important because it represents treatment failure through resistant dominance.
4.3 Absence of A Sensitive-only Equilibrium When Mutation Is Present
A sensitive-only equilibrium would have the form However, if equation (4.3) becomes Therefore, if and this condition cannot be satisfied. Hence, no positive sensitive-only equilibrium exists when mutation from sensitivity to resistance is present.
Proposition 4.1 If the model has no sensitive-only equilibrium with and .
Proof. Assume that and Substituting into the second equilibrium equation gives
Since and this is impossible. Therefore, no positive sensitive-only equilibrium exists when This proves the result.
This result is biologically important. It means that a purely sensitive tumor state is not an equilibrium when mutation or phenotypic switching to resistance is present. Even if the tumor initially consists only of sensitive cells, resistant cells may emerge through the conversion term Thus, resistance-free tumor dynamics are not structurally robust when
4.4 Sensitive-only Equilibrium in The Special Case μ = 0
When the resistant-free boundary becomes invariant. In this special case, a sensitive-only equilibrium may exist. Setting and in equation (4.2), we obtain
For this gives
Therefore, if the sensitive-only equilibrium exists and is given by
This equilibrium exists only when treatment is weaker than the intrinsic growth capacity of sensitive cells. If the positive sensitive-only state is eliminated.
4.5 Coexistence Equilibrium
We now derive the coexistence equilibrium
For equation (4.2) gives
Define
(4.4)
Then
(4.5)
A necessary condition for coexistence is
That is,
(4.6)
If the net growth capacity of sensitive cells is not sufficient to overcome treatment and conversion to resistance. In that case, no coexistence equilibrium with can exist. From equation (4.5), we obtain
(4.7)
Substituting this expression into equation (4.3) gives a quadratic equation for :
(4.8)
where
and
When the possible coexistence values of are
(4.9)
If and equation (4.8) reduces to a linear equation. In this case, the possible coexistence value is
If and equation (4.8) reduces to . This degenerate case is excluded from the generic threshold classification. It would require a separate case analysis.
The corresponding sensitive population is
(4.10)
Therefore, in the quadratic case a biologically feasible coexistence equilibrium exists when and at least one value of in (4.9) satisfies and In the linear case and feasibility requires and If the last condition can be written as If it reduces to since
The coexistence equilibrium represents a mixed tumor state in which sensitive and resistant cells persist together. The condition means that sensitive cells retain positive net growth after treatment and conversion losses. The remaining feasibility conditions determine whether clonal competition permits both populations to coexist.
4.6 Threshold Quantity for Sensitive-cell Persistence
The quantity
is a key threshold for the persistence of sensitive cells. It compares the intrinsic growth rate of sensitive cells with the combined loss due to treatment and conversion to resistance.
If then sensitive cells can persist in principle. This is equivalent to If then sensitive cells cannot maintain positive density in any coexistence equilibrium. We call the sensitive-cell persistence threshold.
The threshold measures whether sensitive cells have enough growth capacity to survive under therapy and conversion to resistance. Increasing decreases . Increasing also decreases . Thus, strong therapy and strong conversion to resistance both reduce the possibility of long-term sensitive-cell persistence.
4.7 Threshold for Invasion of The Resistant-only State by Sensitive Cells
The resistant-only equilibrium is To understand whether sensitive cells can invade this state, we examine the per-capita growth rate of rare sensitive cells near . When S is small and the sensitive equation gives
Define
(4.11)
If then rare sensitive cells can increase when introduced into a resistant-dominated tumor. If then sensitive cells cannot invade the resistant-only state. The case represents the threshold between invasion and non-invasion of sensitive cells near the resistant-only equilibrium.
The quantity measures the ability of sensitive cells to re-enter a resistant-dominated tumor. A large value of means that resistant cells strongly suppress sensitive cells. A large value of means that treatment strongly suppresses sensitive cells. A large value of means that sensitive cells are rapidly converted to resistant cells. All three effects reduce Therefore, when resistant dominance is reinforced by competition, treatment, and conversion.
4.8 Threshold for Resistant-cell Invasion When μ = 0
When resistant cells are continuously generated from sensitive cells. Hence, the resistant-free boundary is not invariant. In that case, resistant emergence does not require invasion from an initially positive resistant population. However, when the resistant-free boundary is invariant. Then it is meaningful to ask whether rare resistant cells can invade a sensitive-only equilibrium. For the sensitive-only equilibrium is
The per-capita growth rate of rare resistant cells near is
Define
(4.12)
If then rare resistant cells can invade the sensitive-only equilibrium. If then sensitive cells suppress resistant cells through competition. The case represents the invasion threshold for resistant cells in the mutation-free case.
The threshold describes resistant-cell invasion in the mutation-free case. A larger strengthens the competitive suppression of resistant cells by sensitive cells. However, increasing lowers the sensitive-cell equilibrium and weakens this suppression. Thus, treatment may indirectly promote resistant invasion through therapy-induced competitive release.
4.9 Summary of Equilibrium Structure
The equilibrium structure of the model can be summarized as follows. The tumor-free equilibrium always exists. The resistant-only equilibrium also always exists. When no positive sensitive-only equilibrium exists. This is because resistant cells are generated from sensitive cells through the conversion term When a sensitive-only equilibrium exists if . It is given by
A coexistence equilibrium may exist when the sensitive-cell persistence threshold satisfies The additional feasibility conditions for and must also hold. The main threshold quantities derived in this section are
and, in the special case
These quantities will be used in the next section to analyze local and global stability. Similar threshold-based interpretations are standard in mathematical ecology and competition models [27]. The thresholds describe how treatment, mutation, and clonal competition shape tumor composition. Increasing treatment intensity lowers sensitive-cell persistence. It also weakens the suppression of resistant cells. This may create ecological opportunity for resistant invasion. The threshold applies only in the mutation-free case, where the sensitive-only equilibrium can exist.
5. LOCAL AND GLOBAL STABILITY
In this section, we analyze the local stability of the equilibria derived in Section 4. We also establish a global convergence result under a biologically relevant condition. The aim is to determine how treatment intensity, mutation, and interclonal competition influence the long-term behavior of the sensitive and resistant tumor populations. The model was formulated in Section 2 and restated as system (3.1) in Section 3. We therefore do not repeat the full system here.
Let
and
The Jacobian matrix of system (3.1) is
(5.1)
This matrix will be evaluated at each equilibrium point. For a two-dimensional system, local stability can be studied using the trace and determinant of the Jacobian matrix [28]. If the trace is negative and the determinant is positive, then the equilibrium is locally asymptotically stable. If the determinant is negative, then the equilibrium is a saddle point.
5.1 Stability of The Tumor-Free Equilibrium
The tumor-free equilibrium is Evaluating the Jacobian matrix at we obtain
The eigenvalues are Since the tumor-free equilibrium is always unstable.
Theorem 5.1 The tumor-free equilibrium is unstable for all admissible parameter values.
Proof. The Jacobian matrix at has eigenvalue Since at least one eigenvalue has positive real part. Therefore, is unstable. This proves the result.
The instability of reflects a limitation of the present model. Since resistant cells are not directly killed by treatment, any small resistant population can grow when introduced into the system. Thus, complete tumor eradication is not a robust outcome in this framework. The tumor-free state is mathematically present, but it is not dynamically stable under resistant-cell growth.
5.2 Stability of The Resistant-only Equilibrium
The resistant-only equilibrium is Evaluating the Jacobian matrix at we obtain
The eigenvalues are and Recall from Section 4 that Thus, the eigenvalues at are
Theorem 5.2 The resistant-only equilibrium is locally asymptotically stable if It is unstable if . If then is non-hyperbolic.
Proof. At , one eigenvalue is . The other eigenvalue is Therefore, both eigenvalues are negative if In this case, is locally asymptotically stable. If then one eigenvalue is positive and the other is negative. Hence, is a saddle point and is unstable. If one eigenvalue is zero. Thus, is non-hyperbolic. This proves the theorem.
The quantity measures whether rare sensitive cells can invade a resistant-dominated tumor. If sensitive cells cannot increase near The resistant-only state is then locally stable. This means that resistant dominance is reinforced by treatment, mutation, and competition from resistant cells. If rare sensitive cells can grow near . In that case, resistant dominance can be locally disrupted.
5.3 Stability of The Sensitive-Only Equilibrium in The Mutation-Free Case
The sensitive-only equilibrium exists only in the special case and In this case,
Let Then Evaluating the Jacobian matrix at with we obtain
The eigenvalues are and Recall that
Since we have
Theorem 5.3 Assume that and The sensitive-only equilibrium is locally asymptotically stable if . It is unstable if If then is non-hyperbolic.
Proof. Since the first eigenvalue satisfies The second eigenvalue is Therefore, both eigenvalues are negative if Hence, is locally asymptotically stable. If then one eigenvalue is positive. Hence, is unstable. If then one eigenvalue is zero, and the equilibrium is non-hyperbolic. This proves the theorem.
The sensitive-only equilibrium is meaningful only when mutation to resistance is absent. The threshold determines whether rare resistant cells can invade this sensitive tumor state. If sensitive cells suppress resistant cells through competition. If resistant cells can invade. Increasing treatment intensity lowers the sensitive equilibrium level. This weakens the competitive pressure imposed on resistant cells. Hence, treatment may unintentionally promote resistant invasion through competitive release.
5.4 Stability of The Coexistence Equilibrium
Let
be a biologically feasible coexistence equilibrium. Using the equilibrium conditions, the diagonal entries of the Jacobian satisfy
and
The trace of the Jacobian at is
The determinant can be written as
(5.2)
Define
(5.3)
Since the sign of det is determined by . Therefore, the coexistence equilibrium is locally asymptotically stable if and is a saddle point if .
Theorem 5.4 Let be a biologically feasible coexistence equilibrium. Then is locally asymptotically stable if It is unstable if If then is non-hyperbolic.
Proof. At , the trace satisfies The determinant satisfies
Since the determinant is positive if Therefore, if the trace is negative and the determinant is positive. Hence, is locally asymptotically stable. If then the determinant is negative. Therefore, is a saddle point and is unstable. If then the determinant is zero. Hence, the equilibrium is non-hyperbolic. This proves the theorem.
The stability of depends on mutation, competition, and resistant-cell abundance. Terms involving represent continual transfer from sensitive to resistant cells. The factor reflects reciprocal competition. When coexistence is more likely to be stable. When competition may destabilize coexistence.
5.5 Global Elimination of Sensitive Cells and Resistant Dominance
We now consider the case in which treatment and conversion exceed the intrinsic growth capacity of sensitive cells. Assume that
(5.4)
This condition means that sensitive cells cannot compensate for treatment and conversion losses. It therefore suggests global elimination of the sensitive population.
Theorem 5.5 Assume that Then, for all non-negative initial conditions. Moreover, if or if and then Hence,
Proof. From the sensitive-cell equation, we have
Since comparison gives
Hence, as By the positivity and boundedness results in Section 3, remains non-negative and bounded. The resistant-cell equation can be written as
Since and is bounded, the second term tends to zero. Thus, the resistant-cell dynamics approach the logistic equation
If then remains positive and converges to . If and then resistant cells are generated through the term Hence, for some and the same conclusion follows. Therefore, and This proves the theorem.
The condition means that sensitive cells cannot compensate for treatment and conversion losses. Thus, the sensitive population is eliminated asymptotically. However, this does not imply tumor eradication. The model predicts a resistant-dominated tumor at carrying capacity, highlighting the risk of strong sensitive-cell suppression.
5.6 Summary of Stability Results
The stability results can be summarized as follows. The tumor-free equilibrium is always unstable. This is because resistant cells have positive intrinsic growth. The resistant-only equilibrium is locally asymptotically stable if It is unstable if
In the mutation-free case the sensitive-only equilibrium
is locally asymptotically stable if It is unstable if A coexistence equilibrium is locally asymptotically stable if It is unstable if
If then sensitive cells are globally eliminated. Under biologically relevant conditions that allow resistant cells to be present or generated, the system converges to These results show that treatment intensity does not only reduce tumor burden. It also changes the competitive balance between sensitive and resistant cells. Increasing can reduce sensitive-cell persistence and promote resistant dominance. This supports the biological idea of therapy-induced competitive release.
The threshold cases and correspond to non-hyperbolic equilibria. They may indicate possible bifurcation points.
6. BIFURCATION AND COMPETITIVE RELEASE
This section examines how treatment intensity and clonal competition alter the qualitative dynamics of the model. The thresholds and
determine persistence conditions, local stability transitions, and candidate bifurcation surfaces [29]. Biologically, these thresholds separate sensitive-dominated, resistant-dominated, and coexistence regimes. For convenience, recall that
and, in the mutation-free case,
The quantity controls the possible persistence of sensitive cells. The quantity determines whether sensitive cells can invade the resistant-only equilibrium. The quantity determines whether resistant cells can invade the sensitive-only equilibrium in the special case
6.1 Treatment Intensity as A Bifurcation Parameter
We first examine how the treatment intensity changes the threshold quantities. Since
we have
Similarly,
Thus, increasing treatment intensity reduces sensitive-cell persistence and makes invasion of the resistant-only equilibrium more difficult for sensitive cells. In the mutation-free case,
Thus, increasing promotes resistant-cell invasion in the mutation-free setting. These relations summarize the mechanism of therapy-induced competitive release. Stronger treatment suppresses sensitive cells and weakens their competitive control over resistant cells.
6.2 Bifurcation at The Resistant-only Equilibrium
The resistant-only equilibrium is From Section 5, the eigenvalues at are
Since the stability of is determined by . The critical condition is . That is,
Solving for the critical treatment intensity gives
When we have and the resistant-only equilibrium is locally asymptotically stable. When we have and the resistant-only equilibrium is unstable.
Proposition 6.1 The resistant-only equilibrium changes stability when
If this threshold separates a regime in which sensitive cells can invade the resistant-only state from a regime in which resistant dominance is locally stable. If then for all admissible If then at and for In this case, the resistant-only equilibrium is locally stable for all non-negative treatment intensities, except possibly at the boundary case when
Proof. At one eigenvalue is always negative The other eigenvalue is
Thus, precisely when
The sign of changes as passes through this value. Therefore, the local stability of changes at . This proves the result.
The threshold when positive, measures the treatment level above which resistant dominance becomes locally stable. A higher treatment intensity suppresses sensitive cells and prevents them from invading a resistant-dominated tumor. Thus, strong therapy can stabilize the resistant-only state. This is a mathematical expression of therapy-induced selection for resistance.
6.3 Loss of Sensitive-cell Persistence
The sensitive-cell persistence threshold is
The critical condition for loss of sensitive-cell persistence is This gives
Solving for we obtain If then and sensitive cells may persist in a coexistence equilibrium. If then and no coexistence equilibrium with can exist.
Proposition 6.2 The condition defines a threshold for the loss of sensitive-cell persistence. If then the model has no biologically feasible coexistence equilibrium with
Proof. From the coexistence condition derived in Section 4, If a biologically feasible coexistence equilibrium exists, then Since the left-hand side satisfies Therefore, we must have Since this requires Thus, if no coexistence equilibrium with can exist. Since is equivalent to the result follows.
If then for all admissible Thus, sensitive-cell persistence is impossible for all non-negative treatment intensities in this parameter regime.
The threshold represents the treatment level at which sensitive cells lose their ability to persist after accounting for both treatment-induced killing and conversion to resistance. If is too large, sensitive cells cannot maintain a positive population. However, this does not imply tumor eradication. Resistant cells may still persist and dominate the tumor population.
6.4 Resistant-cell Invasion and Competitive Release in The Mutation-free Case
We now consider the special case In this case, the resistant-free boundary is invariant. The sensitive-only equilibrium exists when It is given by
The invasion rate of rare resistant cells is
Setting gives the critical treatment intensity
This threshold is biologically meaningful when because then If then and rare resistant cells cannot invade the sensitive-only equilibrium. If then and resistant cells can invade.
Proposition 6.3 Assume that and . Then resistant cells can invade the sensitive-only equilibrium when
Since we have so the threshold lies within the existence region of Resistant cells cannot invade when
Proof. At the sensitive-only equilibrium, the invasion rate of rare resistant cells is
Since the sign of is determined by the bracketed term. Setting this term equal to zero gives
Thus, when and when This proves the result.
This proposition gives a threshold condition for competitive release. Weak treatment allows sensitive cells to suppress resistant cells. Stronger treatment reduces this suppression and may permit resistant invasion.
6.5 Stability Transition of The Coexistence Equilibrium
Let be a biologically feasible coexistence equilibrium. From Section 5, the stability of is determined by
The coexistence equilibrium is locally asymptotically stable if It is unstable if The critical condition is This condition defines a candidate bifurcation surface for the coexistence equilibrium.
Proposition 6.4 A biologically feasible coexistence equilibrium may change stability when At this threshold, is non-hyperbolic.
Proof. From Section 5, the trace of the Jacobian matrix at is negative, and the determinant is
Since and the sign of is determined by If then
Therefore, at least one eigenvalue is zero. Hence, the coexistence equilibrium is non-hyperbolic. This proves the result.
The condition marks a transition in the stability of the mixed tumor state. If sensitive and resistant cells can coexist locally. If coexistence becomes unstable. This may lead to dominance by one phenotype. The term shows that reciprocal competition plays an important role. Strong reciprocal competition may destabilize coexistence, while weaker reciprocal competition may allow stable mixed tumors.
6.6 Summary of Bifurcation Thresholds
The bifurcation and competitive release structure of the model is governed by several threshold quantities. First, the resistant-only equilibrium changes stability when or equivalently, Second, sensitive-cell persistence is lost when or equivalently, Third, in the mutation-free case, resistant-cell invasion of the sensitive-only equilibrium occurs when or equivalently,
Fourth, the coexistence equilibrium may change local stability when .
The main conclusion of this section is that treatment can have two opposing ecological effects. It suppresses therapy-sensitive cells directly. At the same time, it may reduce competition against resistant cells. This can promote resistant-cell expansion even when the resistant population has no direct growth advantage under low treatment. Therefore, the model supports the idea that aggressive therapy may unintentionally accelerate resistant dominance through competitive release.
7. NUMERICAL SIMULATIONS
In this section, we present numerical simulations to illustrate the analytical results obtained in Sections 4–6. The main aim is to show how treatment intensity, mutation, and clonal competition influence the long-term composition of the tumor. We also illustrate therapy-induced competitive release. Related model-based simulations have also been used to study qualitative dynamical behavior in geospatial-temporal SEIR frameworks [30]. The simulations are not intended to fit a specific clinical dataset. Instead, they are used to demonstrate qualitative dynamical regimes predicted by the model. The simulations are performed using the dimension less form of the model, where the treatment intensity is scaled as :
(7.1)
Here, The dimensionless parameters are
Here, and represent the normalized therapy-sensitive and therapy-resistant tumor populations, respectively. The parameter represents the relative treatment intensity. The parameter represents the relative conversion rate from sensitivity to resistance. The parameter represents the growth rate of resistant cells relative to sensitive cells. The total normalized tumor burden is defined as The resistant fraction is defined as
(7.2)
This fraction is used to quantify resistant dominance. It is well defined whenever
7.1 Numerical Method
System (7.1) is solved numerically using MATLAB’s ode45 solver with stringent relative and absolute tolerances [31]. The relative and absolute tolerances are set to and respectively. The numerical implementation follows standard practices for scientific computing and numerical experiments [32]. The simulations are carried out on the time interval The initial condition is chosen as
unless otherwise stated. This initial condition represents a tumor that is initially dominated by therapy-sensitive cells, with a small resistant subpopulation. For comparison across different parameter regimes, we use the same initial condition in most simulations. This allows the effect of treatment intensity and clonal competition to be isolated more clearly.
The initial condition reflects a clinically relevant situation in which resistant cells are rare but not absent. This may represent pre-existing resistance or early emergence of resistance. Studying this case is important because small resistant populations may expand after therapy reduces the sensitive population.
We first consider a weak-treatment regime. The parameter values are
For these values,
Thus, sensitive cells can persist in principle. The threshold is positive. Hence, the resistant-only equilibrium is not locally stable. The analytical results therefore suggest that the system may approach a coexistence equilibrium rather than resistant dominance.
The numerical solution confirms this behavior. The sensitive population initially increases slightly and then relaxes to a positive steady-state level. The resistant population increases from its small initial value. Both populations eventually approach positive steady-state levels.
Weak therapy does not eliminate the sensitive population. In this parameter regime, sensitive cells remain sufficiently abundant to compete with resistant cells. As a result, the system approaches a mixed tumor state. This regime corresponds to coexistence between sensitive and resistant cell populations.
7.3 Scenario II: Strong Therapy and Resistant Dominance
We next consider a strong-treatment regime. The parameter values are
In this case, Thus,
The analytical result in Section 5 predicts that sensitive cells cannot persist when the combined loss due to treatment and conversion exceeds their intrinsic growth capacity. Therefore, the sensitive population should approach zero. If resistant cells are initially present, or if they are generated through mutation, the resistant population should approach the resistant-only equilibrium. The numerical simulation supports this prediction. The sensitive population declines rapidly. The resistant population increases and approaches the carrying capacity in normalized units.
Under strong therapy, the sensitive population declines rapidly. The total tumor burden may initially decrease. However, resistant cells can expand and occupy the available ecological space. Thus, treatment does not eradicate the tumor in this model. Instead, it may lead to long-term resistant dominance.
7.4 Scenario III: Competitive Release in The Mutation-Free Case
We now examine competitive release in the special case In this case, resistant cells are not generated from sensitive cells. Therefore, resistant invasion depends on whether rare resistant cells can grow when introduced into a sensitive-dominated tumor. We set
The resistant-cell invasion threshold is Since we have Two treatment intensities are compared:
and
When , rare resistant cells cannot invade. Sensitive cells remain abundant and suppress resistant cells through competition. When , rare resistant cells can invade. Treatment reduces the sensitive population and weakens competitive suppression.
This comparison separates the effect of treatment from the effect of mutation. Since resistant cells are not continuously produced from sensitive cells. Therefore, any increase in the resistant population must come from the growth of the initially rare resistant subpopulation. This makes the competitive-release mechanism clearer. In the low-treatment case, sensitive cells remain close to their positive equilibrium and continue to suppress resistant cells. In the high-treatment case, sensitive cells are reduced. The resistant population then faces weaker competition and can expand. Thus, the threshold identifies when treatment changes from suppressing tumor burden to promoting resistant invasion.
This simulation directly illustrates therapy-induced competitive release. When treatment is weak, sensitive cells suppress resistant cells. When treatment is stronger, sensitive cells are reduced. This opens ecological space for resistant cells. Thus, treatment can indirectly promote resistance even when resistance is not continuously generated by mutation.
7.5 Bifurcation Diagram with Respect to Treatment Intensity
To visualize how treatment intensity changes long-term tumor composition, we compute numerical equilibria for a range of values. The parameters
are fixed. The treatment intensity is varied over the interval
For each value of system (7.1) is integrated up to The terminal values are used as numerical approximations of the steady-state value. We then record the limiting values of
The dashed vertical lines in Figure 4 indicate two analytical thresholds, and . The first threshold marks the local stability transition of the resistant-only equilibrium. The second marks the loss of sensitive-cell persistence. Thus, resistant dominance begins before sensitive-cell persistence is lost. At low treatment intensity, sensitive cells remain dominant or coexist with resistant cells. As increases, the sensitive steady state decreases and the resistant fraction increases. Hence, treatment intensity can reshape tumor composition from a mixed state toward resistant dominance.
7.6 Resistant Fraction as A Function of Treatment and Competition
We also examine how resistant dominance depends jointly on treatment intensity and clonal competition. The parameters
are fixed. The treatment intensity and competition coefficient are varied over the ranges and For each pair , system (7.1) is solved numerically. The final resistant fraction
is recorded at the final simulation time
This heat map illustrates the joint effect of therapy and competition. High treatment intensity increases the resistant fraction. Strong competitive pressure from sensitive cells can suppress resistant cells when treatment is low. However, when treatment reduces the sensitive population, this suppressive effect weakens. Therefore, resistant dominance is promoted by the combined effect of strong therapy and reduced sensitive-cell competition.
7.7 Summary of Numerical Findings
The numerical simulations support the analytical results from Sections 4–6. Weak therapy may allow coexistence between sensitive and resistant cells. Strong therapy can eliminate sensitive cells but may lead to resistant dominance. In the mutation-free case, resistant invasion depends on whether treatment reduces sensitive-cell competition below a critical level. This provides a clear numerical illustration of therapy-induced competitive release. Overall, the simulations show that treatment suppresses sensitive cells but also changes ecological balance. When sensitive cells are strongly suppressed, resistant cells may gain access to space and resources. Thus, aggressive treatment may unintentionally promote resistant dominance through competitive release. The simulations illustrate qualitative transitions rather than patient-specific predictions.
8. DISCUSSION AND BIOLOGICAL IMPLICATIONS
This study analyzed a nonlinear model for therapy-sensitive and therapy-resistant tumor populations under clonal competition. The model captures coexistence, resistant dominance, loss of sensitive-cell persistence, and therapy-induced competitive release. It is biologically feasible because solutions remain non-negative and bounded. Thus, tumor trajectories stay within a finite admissible region. These properties justify the equilibrium, stability, and bifurcation analyses used to interpret long-term tumor dynamics.
The equilibrium analysis shows that resistant dominance is one important long-term outcome. The resistant-only equilibrium represents a tumor state in which sensitive cells have disappeared and resistant cells occupy the available ecological space. This state corresponds to treatment failure through resistance. The threshold quantities clarify this mechanism. The sensitive-cell persistence threshold determines whether sensitive cells can remain positive under treatment and conversion. If treatment intensity or conversion is too high, sensitive cells cannot persist. The resistant population may then remain and dominate. Thus, strong therapy can remove the sensitive compartment without eradicating the tumor.
The stability results show that the resistant-only equilibrium can become locally stable when sensitive cells cannot invade a resistant-dominated tumor. Thus, therapy is not only cytotoxic. It also acts as a selective pressure that changes tumor composition [1, 2, 5–7]. This leads to therapy-induced competitive release. Under weak treatment, sensitive cells may suppress resistant cells through competition. Under stronger treatment, sensitive cells decline, and this suppression weakens. Resistant cells may then expand. This mechanism is consistent with adaptive therapy. In that strategy, preserving sensitive cells can help control resistance [12–15].
The numerical simulations support the analytical results. Under weak therapy, sensitive and resistant cells approach coexistence. Under strong therapy, sensitive cells decline and resistant cells expand. The total tumor burden may decrease initially but later rebound as resistant cells occupy available ecological space. Thus, short-term tumor reduction does not necessarily imply long-term control. The bifurcation diagram and heat map show that treatment intensity reshapes tumor composition. As treatment increases, the sensitive steady state decreases and the resistant fraction increases across a wide range of competition strengths.
These findings have implications for treatment design. Aggressive therapy may reduce sensitive cells but also create ecological space for resistant cells. This supports adaptive therapy, where the goal is not always maximum tumor reduction but maintaining enough sensitive cells to suppress resistance [12–15]. The model also has limitations. Treatment is constant, resistance is complete, and spatial structure is not included. In practice, treatment may vary over time, resistance may be partial, and spatial heterogeneity may affect competition and resistance [16–18]. Thus, the simulations illustrate qualitative regimes rather than patient-specific predictions.
Future work should replace the constant treatment intensity with a control function to study adaptive schedules. Parameter estimation from experimental or clinical data is another important extension. This would allow the threshold quantities to be linked more directly to observed tumor dynamics. The model may also include partial resistance, treatment-dependent switching, time delay, or spatial diffusion. Taken together, the results show that resistant dominance can emerge from treatment pressure and clonal competition through therapy-induced competitive release.
9. CONCLUSIONS
This study proposed and analyzed a nonlinear two-population model for therapy-sensitive and therapy-resistant tumor cells. The model includes logistic growth, asymmetric clonal competition, conversion from sensitivity to resistance, and treatment acting directly on sensitive cells. The aim was to examine how treatment pressure changes the ecological balance between the two tumor phenotypes.
The analysis showed that the model is biologically feasible. Solutions remain non-negative and bounded for non-negative initial data. Hence, the model produces meaningful tumor trajectories over long time intervals. The equilibrium analysis identified several admissible steady states. These include the tumor-free state, resistant-only dominance, sensitive-only persistence in the mutation-free case, and coexistence of sensitive and resistant cells. The tumor-free state exists mathematically, but it is not stable in the present model.
The threshold quantities provide a useful interpretation of therapy-induced selection. They determine whether sensitive cells persist, whether resistant dominance is stable, and whether rare resistant cells can invade in the mutation-free case. These thresholds connect treatment intensity, mutation, and clonal competition to qualitative tumor outcomes.
The stability, bifurcation, and numerical results show that increasing treatment intensity can promote resistant dominance. Strong treatment suppresses sensitive cells directly. However, it also weakens the competitive pressure that sensitive cells exert on resistant cells. This creates ecological space for resistant expansion. Thus, treatment may reduce tumor burden in the short term but favor resistance in the long term.
These findings indicate that treatment intensity should not be viewed only as a tumor-reducing factor. It also acts as a selective ecological force. Future studies should consider time-dependent treatment, partial resistance, spatial diffusion, and data-informed parameter estimation.
ACKNOWLEDGEMENTS
The author would like to thank the Department of Mathematics, Faculty of Science, Srinakharinwirot University, for its academic support.
AUTHOR CONTRIBUTIONS
Panyawat Haarsa: Conceptualization, Methodology, Software, Formal analysis, Investigation, Visualization, Writing – original draft, Writing – review & editing.
CONFLICT OF INTEREST STATEMENT
The author declares no conflicts of interest.
DECLARATION OF USE OF GENERATIVE AI
During the preparation of this manuscript, generative AI tools were used to enhance the clarity and readability of the language. The manuscript was carefully reviewed and revised, and full responsibility for the content is retained by the author.
FUNDING
This research received no external funding.
REFERENCES
[1] Nowell P.C., The clonal evolution of tumor cell populations. Science, 1976; 194(4260): 23-28. DOI 10.1126/science.959840.
[2] Merlo L.M.F., Pepper J.W., Reid B.J. and Maley C.C., Cancer as an evolutionary and ecological process. Nature Reviews Cancer, 2006; 6(12): 924-935. DOI 10.1038/nrc2013.
[3] Kareva I., Biological stoichiometry in tumor micro-environments. PLoS One, 2013; 8(1): e51844. DOI 10.1371/journal.pone.0051844.
[4] Maley C.C., Aktipis A., Graham T.A., Sottoriva A., Boddy A.M., Janiszewska M., et al., Classifying the evolutionary and ecological features of neoplasms. Nature Reviews Cancer, 2017; 17(10): 605-619. DOI 10.1038/nrc.2017.69.
[5] Greaves M. and Maley C.C., Clonal evolution in cancer. Nature, 2012; 481(7381): 306-313. DOI 10.1038/nature10762.
[6] Holohan C., Van Schaeybroeck S., Longley D.B. and Johnston P.G., Cancer drug resistance: An evolving paradigm. Nature Reviews Cancer, 2013; 13(10): 714-726. DOI 10.1038/nrc3599.
[7] Huijben S., Bell A.S., Sim D.G., Tomasello D., Mideo N., Day T., et al., Aggressive chemotherapy and the selection of drug resistant pathogens. PLoS Pathogens, 2013; 9(9): e1003578. DOI 10.1371/journal.ppat.1003578.
[8] Azizi T., Mathematical modelling of cancer treatments, resistance, optimization. AppliedMath, 2025; 5(2): 40. DOI 10.3390/appliedmath5020040.
[9] Altrock P.M., Liu L.L. and Michor F., The mathematics of cancer: Integrating quantitative models. Nature Reviews Cancer, 2015; 15(12): 730-745. DOI 10.1038/nrc4029.
[10] Yin A., Moes D.J.A.R., van Hasselt J.G.C., Swen J.J. and Guchelaar H.-J., A review of mathematical models for tumor dynamics and treatment resistance evolution of solid tumors. CPT: Pharmacometrics & Systems Pharmacology, 2019; 8(10): 720-737. DOI 10.1002/psp4.12450.
[11] Anderson A.R.A. and Quaranta V., Integrative mathematical oncology. Nature Reviews Cancer, 2008; 8(3): 227-234. DOI 10.1038/nrc2329.
[12] Gatenby R.A., Silva A.S., Gillies R.J. and Frieden B.R., Adaptive therapy. Cancer Research, 2009; 69(11): 4894-4903. DOI 10.1158/0008-5472.CAN-08-3658.
[13] Wodarz D., Adaptive therapy and the cost of drug-resistant mutants. Cancer Research, 2021; 81(4): 811-812. DOI 10.1158/0008-5472.CAN-20-4079.
[14] Stanková K., Brown J.S., Dalton W.S. and Gatenby R.A., Optimizing cancer treatment using game theory: A review. JAMA Oncology, 2019; 5(1): 96-103. DOI 10.1001/jamaoncol.2018.3395.
[15] Gallagher K., Strobl M., Anderson A.R.A. and Maini P.K., Deriving optimal treatment timing for adaptive therapy: Matching the model to the tumor dynamics. Bulletin of Mathematical Biology, 2025; 87: 146. DOI 10.1007/s11538-025-01525-y.
[16] Bacevic K., Noble R., Soffar A., Ammar O.W., Boszonyik B., Prieto S., et al., Spatial competition constrains resistance to targeted cancer therapy. Nature Communications, 2017; 8: 1995. DOI 10.1038/s41467-017-01516-1.
[17] Strobl M.A.R., Gallaher J., West J., Robertson-Tessi M., Maini P.K. and Anderson A.R.A., Spatial structure impacts adaptive therapy by shaping intra-tumoral competition. Communications Medicine, 2022; 2: 46. DOI 10.1038/s43856-022-00110-x.
[18] Gallaher J., Enriquez-Navas P.M., Luddy K.A., Gatenby R.A. and Anderson A.R.A., Spatial heterogeneity and evolutionary dynamics modulate time to recurrence in continuous and adaptive cancer therapies. Cancer Research, 2018; 78(8): 2127-2139. DOI 10.1158/0008-5472.CAN-17-2649.
[19] McGehee C. and Mori Y., A mathematical framework for comparison of intermittent versus continuous adaptive chemotherapy dosing in cancer. npj Systems Biology and Applications, 2024; 10: 140. DOI 10.1038/s41540-024-00461-2.
[20] Laird A.K., Dynamics of tumor growth. British Journal of Cancer, 1964; 18: 490-502. DOI 10.1038/bjc.1964.55.
[21] Murray J.D., Mathematical Biology I: An Introduction, 3rd Edn., Springer, New York, 2002.
[22] Komarova N.L. and Wodarz D., Drug resistance in cancer: Principles of emergence and prevention. Proceedings of the National Academy of Sciences, 2005; 102(27): 9714-9719. DOI 10.1073/pnas.0501870102.
[23] Hale J.K., Ordinary Differential Equations, Wiley-Interscience, New York, 1969.
[24] Perko L., Differential Equations and Dynamical Systems, 3rd Edn., Springer, New York, 2001.
[25] Smith H.L. and Waltman P., The Theory of the Chemostat: Dynamics of Microbial Competition, Cambridge University Press, Cambridge, 1995.
[26] West J., Adler F., Gallaher J., Strobl M., Brady-Nicholls R., Brown J., et al., A survey of open questions in adaptive therapy: Bridging mathematics and clinical translation. eLife, 2023; 12: e84263. DOI 10.7554/eLife.84263.
[27] Kot M., Elements of Mathematical Ecology, Cambridge University Press, Cambridge, 2001.
[28] Hirsch M.W., Smale S. and Devaney R.L., Differential Equations, Dynamical Systems, and an Introduction to Chaos, 3rd Edn., Academic Press, Amsterdam, 2013.
[29] Kuznetsov Y.A., Elements of Applied Bifurcation Theory, 3rd Edn., Springer Cham, New York, 2004.
[30] Punzal C.D.F. and Tapang G.A., Epidemic dynamics on a geospatial-temporal network using a mobility-informed SEIR model. Chiang Mai Journal of Science, 2026; 53(2): e2026031. DOI 10.12982/CMJS.2026.031.
[31] Hairer E., Nørsett S.P. and Wanner G., Solving Ordinary Differential Equations I: Nonstiff Problems, 2nd Edn., Springer, Berlin, 1993.
[32] Press W.H., Teukolsky S.A., Vetterling S.P. and Flannery B.P., Numerical Recipes: The Art of Scientific Computing, 3rd Edn., Cambridge University Press, Cambridge, 2007.