Processing math: 52%
Research article

DTiGNN: Learning drug-target embedding from a heterogeneous biological network based on a two-level attention-based graph neural network


  • Motivation: In vitro experiment-based drug-target interaction (DTI) exploration demands more human, financial and data resources. In silico approaches have been recommended for predicting DTIs to reduce time and cost. During the drug development process, one can analyze the therapeutic effect of the drug for a particular disease by identifying how the drug binds to the target for treating that disease. Hence, DTI plays a major role in drug discovery. Many computational methods have been developed for DTI prediction. However, the existing methods have limitations in terms of capturing the interactions via multiple semantics between drug and target nodes in a heterogeneous biological network (HBN). Methods: In this paper, we propose a DTiGNN framework for identifying unknown drug-target pairs. The DTiGNN first calculates the similarity between the drug and target from multiple perspectives. Then, the features of drugs and targets from each perspective are learned separately by using a novel method termed an information entropy-based random walk. Next, all of the learned features from different perspectives are integrated into a single drug and target similarity network by using a multi-view convolutional neural network. Using the integrated similarity networks, drug interactions, drug-disease associations, protein interactions and protein-disease association, the HBN is constructed. Next, a novel embedding algorithm called a meta-graph guided graph neural network is used to learn the embedding of drugs and targets. Then, a convolutional neural network is employed to infer new DTIs after balancing the sample using oversampling techniques. Results: The DTiGNN is applied to various datasets, and the result shows better performance in terms of the area under receiver operating characteristic curve (AUC) and area under precision-recall curve (AUPR), with scores of 0.98 and 0.99, respectively. There are 23,739 newly predicted DTI pairs in total.

    Citation: Saranya Muniyappan, Arockia Xavier Annie Rayan, Geetha Thekkumpurath Varrieth. DTiGNN: Learning drug-target embedding from a heterogeneous biological network based on a two-level attention-based graph neural network[J]. Mathematical Biosciences and Engineering, 2023, 20(5): 9530-9571. doi: 10.3934/mbe.2023419

    Related Papers:

    [1] Yong Luo . Global existence and stability of the classical solution to a density-dependent prey-predator model with indirect prey-taxis. Mathematical Biosciences and Engineering, 2021, 18(5): 6672-6699. doi: 10.3934/mbe.2021331
    [2] Yongli Cai, Malay Banerjee, Yun Kang, Weiming Wang . Spatiotemporal complexity in a predator--prey model with weak Allee effects. Mathematical Biosciences and Engineering, 2014, 11(6): 1247-1274. doi: 10.3934/mbe.2014.11.1247
    [3] Paulo Amorim, Bruno Telch, Luis M. Villada . A reaction-diffusion predator-prey model with pursuit, evasion, and nonlocal sensing. Mathematical Biosciences and Engineering, 2019, 16(5): 5114-5145. doi: 10.3934/mbe.2019257
    [4] Yuxuan Zhang, Xinmiao Rong, Jimin Zhang . A diffusive predator-prey system with prey refuge and predator cannibalism. Mathematical Biosciences and Engineering, 2019, 16(3): 1445-1470. doi: 10.3934/mbe.2019070
    [5] Raimund Bürger, Gerardo Chowell, Elvis Gavilán, Pep Mulet, Luis M. Villada . Numerical solution of a spatio-temporal predator-prey model with infected prey. Mathematical Biosciences and Engineering, 2019, 16(1): 438-473. doi: 10.3934/mbe.2019021
    [6] Tingfu Feng, Leyun Wu . Global dynamics and pattern formation for predator-prey system with density-dependent motion. Mathematical Biosciences and Engineering, 2023, 20(2): 2296-2320. doi: 10.3934/mbe.2023108
    [7] Jin Zhong, Yue Xia, Lijuan Chen, Fengde Chen . Dynamical analysis of a predator-prey system with fear-induced dispersal between patches. Mathematical Biosciences and Engineering, 2025, 22(5): 1159-1184. doi: 10.3934/mbe.2025042
    [8] Ilse Domínguez-Alemán, Itzel Domínguez-Alemán, Juan Carlos Hernández-Gómez, Francisco J. Ariza-Hernández . A predator-prey fractional model with disease in the prey species. Mathematical Biosciences and Engineering, 2024, 21(3): 3713-3741. doi: 10.3934/mbe.2024164
    [9] Muhammad Shoaib Arif, Kamaleldin Abodayeh, Asad Ejaz . On the stability of the diffusive and non-diffusive predator-prey system with consuming resources and disease in prey species. Mathematical Biosciences and Engineering, 2023, 20(3): 5066-5093. doi: 10.3934/mbe.2023235
    [10] Yuanshi Wang, Donald L. DeAngelis . A mutualism-parasitism system modeling host and parasite with mutualism at low density. Mathematical Biosciences and Engineering, 2012, 9(2): 431-444. doi: 10.3934/mbe.2012.9.431
  • Motivation: In vitro experiment-based drug-target interaction (DTI) exploration demands more human, financial and data resources. In silico approaches have been recommended for predicting DTIs to reduce time and cost. During the drug development process, one can analyze the therapeutic effect of the drug for a particular disease by identifying how the drug binds to the target for treating that disease. Hence, DTI plays a major role in drug discovery. Many computational methods have been developed for DTI prediction. However, the existing methods have limitations in terms of capturing the interactions via multiple semantics between drug and target nodes in a heterogeneous biological network (HBN). Methods: In this paper, we propose a DTiGNN framework for identifying unknown drug-target pairs. The DTiGNN first calculates the similarity between the drug and target from multiple perspectives. Then, the features of drugs and targets from each perspective are learned separately by using a novel method termed an information entropy-based random walk. Next, all of the learned features from different perspectives are integrated into a single drug and target similarity network by using a multi-view convolutional neural network. Using the integrated similarity networks, drug interactions, drug-disease associations, protein interactions and protein-disease association, the HBN is constructed. Next, a novel embedding algorithm called a meta-graph guided graph neural network is used to learn the embedding of drugs and targets. Then, a convolutional neural network is employed to infer new DTIs after balancing the sample using oversampling techniques. Results: The DTiGNN is applied to various datasets, and the result shows better performance in terms of the area under receiver operating characteristic curve (AUC) and area under precision-recall curve (AUPR), with scores of 0.98 and 0.99, respectively. There are 23,739 newly predicted DTI pairs in total.



    In 1987, Karevia and Odell [1] first proposed the following one-predator and one-prey model with prey-taxis in order to explain that an area-restricted search creates the following predator aggregation phenomenon

    {ut=(d(w)u)(uχ(w)w)+G1(u,w),wt=DΔw+G2(u,w), (1.1)

    where D>0 is the diffusivity coefficient of preys, d(w) denotes the motility function of predators, χ(w) represents the prey-taxis sensitivity coefficient, and the term (uχ(w)w) stands for the tendency of the predator moving towards the increasing direction of the prey gradient, and it is viewed as the prey-taxis term. The functions G1(u,w) and G2(u,w) describe the predator-prey interactions, which include both intra-specific and inter-specific interactions. Generally the predator-prey interaction functions G1(u,w) and G2(u,w) possess the following prototypical forms.

    G1(u,w)=γuF(w)uh(u),G2(u,w)=uF(w)+f(w), (1.2)

    where γ>0 denotes the intrinsic predation rate, uF(w) represents the inter-specific interaction and uh(u) and f(w) stand for the intra-specific interaction. Specifically, F(w) is the functional response function accounting for the intake rate of predators as a function of prey density; it is often used in the following form in the literature [2,3,4]

    F(w)=w(Holling type I),F(w)=wλ+w(Holling type II),F(w)=wmλm+wm(Holling type III),F(w)=1eλw(Ivlev type) (1.3)

    with constants λ>0 and m>1; other types of functional response functions (e.g., Beddington-DeAngelis type in [5], Crowley-Martin type in [6]) and more predator-prey interactions can be found in [7,8,9,10]. The predator mortality rate function h(u) is typically of the form

    h(u)=θ+αu, (1.4)

    where θ>0 accounts for the natural death rate and α0 denotes the rate of death resulting from the intra-specific competition, which is also called the density-dependent death [11]. The prey growth function f(w) is usually assumed to be negative for large w due to the limitation of resources (or crowding effect), and its typical forms are

    f(w)=μw(1wK)(Logistic type),orf(w)=μw(1wK)(wk1)(Bistable or Allee effect type), (1.5)

    where μ>0 is the intrinsic growth rate of prey, K>0 is called the carrying capacity and 0<k<K. Now there exist many interesting results about global existence, uniform boundedness, asymptotic behavior, traveling waves and pattern formation of solutions to System (1.1) or its variants in [11,12,13,14,15]. When d(w)=d>0 and χ(w)=χ>0, G1(u,v) and G2(u,v) have the forms of (1.2), Wu et al. [15] obtained the global existence and uniform persistence of solutions to (1.1) in any dimension provided that χ is suitably small. Then, Jin and Wang[13] derived the global boundedness and asymptotic stability of solutions for System (1.1) without the smallness assumption on χ in a two-dimensional bounded domain. Moreover, Wang et al. [16] studied the nonconstant positive steady states and pattern formation of (1.1) in a one-dimensional bounded domain. Under the conditions that d(w) and χ(w) are not constants and h(u) is given by (1.4), Jin and Wang[14] established the global boundedness, asymptotic behavior and spatio-temporal patterns of solutions for (1.1) under some conditions on the parameters in a two-dimensional smooth bounded domain. For more related results in predator-prey models, we refer the readers to [17,18,19,20,21,22,23,24,25,26,27,28,29] and the references therein.

    However, all the aforementioned works are devoted to studying prey-taxis models with one-predator and one prey. Now let us mention some predator-prey models with two-predator and one-prey. Recently, the following general two-predator and one-prey model with prey-taxis has attracted a lot of attention.

    {ut=(d1(w)u)(uχ1(w)w)+γ1uF1(w)uh1(u)β1uv,vt=(d2(w)v)(vχ2(w)w)+γ2vF2(w)vh2(v)β2uv,wt=DΔwuF1(w)vF2(w)+f(w), (1.6)

    as applied in a smooth bounded domain ΩRn, n1. Given d1(w)=d2(w)=1 and β1=β2=0, Wang et al. [30] derived the global boundedness, nonconstant positive steady states and time-periodic patterns of solutions for System (1.6). Wang and Wang [31] studied the uniform boundedness and asymptotic stability of nonnegative spatially homogeneous equilibria for (1.6) in any dimension. Given d1(w)=d2(w)=1, χi(w)=χi>0(i=1,2) and β1=β2=β>0, Mi et al. [32] obtained the global boundedness and stability of classical solutions in any dimension under suitable conditions of parameters. Under the conditions that d1(w) and d2(w) are non-constants and β1=β2=0, Qiu et al.[33] rigorously proved the global existence, uniform boundedness and stabilization of classical solutions in any dimension with suitable conditions on motility functions and the coefficients of the logistic source. However, when β1,β20, the global existence and stabilization of solutions for (1.6) are still open. Given χi(w)=di(w)0 if di(w)0(i=1,2), the diffusion-advection terms in (1.6) can respectively become the forms Δ(d1(w)u) and Δ(d2(w)v), which could be interpreted as "density-suppressed motility" in [34,35]. This means that the predator will reduce its motility when encountering the prey, which is a rather reasonable assumption that has very sound applications in the predator-prey systems. Since the possible degeneracy caused by the density-suppressed motility brings considerable challenges for mathematical analysis, many works have showed various interesting results, which can be found in [36,37,38,39,40,41,42,43,44]. Given χi(w)=di(w)0(i=1,2), F1(w)=F2(w)=w, h1(u)=u,h2(v)=v and f(w)=μw(m(x)w), System (1.6) can be simplified as

    {ut=Δ(d1(w)u)+u(γ1wuβ1v),vt=Δ(d2(w)v)+v(γ2wvβ2u),wt=DΔw(u+v)w+μw(m(x)w), (1.7)

    where the parameters D,μ,γi,βi(i=1,2) are positive, the dispersal rate functions di(w)(i=1,2) satisfy following the hypothesis: di(w)C2([0,)),di(w)0 on [0,) and di(w)>0. Wang and Xu [45] have found some interesting results for System (1.7) in a two-dimensional smooth bounded domain. More specifically, when D=1 and m(x)=1, System (1.7) has a unique globally bounded classical solution. By constructing appropriate Lyapunov functionals and using LaSalle's invariant principle, the authors proved that the global bounded solution of (1.7) converges to the co-existence steady state exponentially or competitive exclusion steady state algebraically as time tends to infinity in different parameter regimes. For a prey's resource that is spatially heterogeneous (i.e., m(x) is non-constant), the authors used numerical simulations to demonstrate that the striking phenomenon "slower diffuser always prevails" given in [46,47] fails to appear if the non-random dispersal strategy is employed by competing species (i.e., either d1(w) or d2(w) is non-constant) while it still holds if both d1(w) and d2(w) are constants. However, there are few results about global boundedness and large time behavior of solutions for (1.7) in the general form.

    Inspired by the above works, this paper is concerned with the following two-species competitive predator-prey system with the following density-dependent diffusion

    {ut=Δ(d1(w)u)+γ1uF1(w)uh1(u)β1uv,(x,t)Ω×(0,),vt=Δ(d2(w)v)+γ2vF2(w)vh2(v)β2uv,(x,t)Ω×(0,),wt=DΔwuF1(w)vF2(w)+f(w),(x,t)Ω×(0,),uν=vν=wν=0,(x,t)Ω×(0,),(u,v,w)(x,0)=(u0,v0,w0)(x),xΩ, (1.8)

    where ΩR2 is a bounded domain with a smooth boundary Ω, ν denotes the derivative with respect to the outward normal vector ν of Ω, and the parameters D and γi, βi (i=1,2) are positive. The unknown functions u=u(x,t) and v=v(x,t) denote the densities of two-competing species (e.g., predators), and w=w(x,t) represents the density of predators' resources (e.g., the prey) at a position x and time t>0. When d1(w)=d1>0 and d2(w)=d2>0, System (1.8) becomes the well-known diffusive predator-prey system, which has been widely studied in [48,49,50,51]. However, to the best of our knowledge, the results of the two-predator and one-prey system given by (1.8) with density-suppressed motility (i.e., d1(w) and d2(w) are non-constants) indicate that the competition and general predator mortality rate hi(u) are almost vacant. The main aim of this paper is to explore the influence of the predation interaction, competition and general predator mortality on the dynamical behavior of System (1.8). Throughout this paper, we assume that the functions di(s),Fi(s),hi(s),(i=1,2),f(s) and initial data (u0,v0,w0) mentioned in (1.8) satisfy the following hypotheses:

    (H1) di(s)C2([0,)) with di(s)>0 and di(s)0, i=1,2 on [0,). The typical example is di(s)=1(1+s)κi or di(s)=exp(κis) with κi>0, i=1,2.

    (H2) Fi(s)C1([0,)), Fi(0)=0, Fi(s)>0 and Fi(s)>0, i=1,2 in (0,).

    (H3) hi(s)C2([0,)) and there exist constants θi>0 and αi0 such that hi(s)θi and hi(s)αi, i=1,2 for all s>0.

    (H4) f(s)C1([0,)) with f(0)=0, and there exist positive constants μ and K such that f(s)μs for all s0, f(K)=0 and f(s)<0 for s>K.

    (H5) (u0,v0,w0)(W1,p(Ω))3 with p>2 and u0,v0,w00.

    Here, we note that there are many candidates for the above functions Fi(s), hi(s) and f(s) as in (1.3)–(1.5). Due to the presence of the prey's density dependent diffusion coefficient, Model (1.8) is a cross-diffusion system and the parabolic comparison principle is no longer applicable. Moreover, when α1=α2=0, the key L2-spatiotemporal estimates of u and v cannot be directly derived; thus, the uniform boundedness of solutions is not an obvious result and needs to be justified. Based on the above hypotheses, the first main result of this paper asserts the global existence and boundedness of solutions for System (1.8) as follows.

    Theorem 1.1. Let D,γi,βi>0(i=1,2), ΩR2 be a smooth bounded domain and the hypotheses (H1)–(H5) hold. Then System (1.8) has a unique global nonnegative classical solution (u,v,w)(C(¯Ω×[0,))C2,1(¯Ω×(0,)))3, which is uniformly bounded in time, i.e., there exists a constant C>0 independent of t such that

    u(,t)L(Ω)+v(,t)L(Ω)+w(,t)W1,(Ω)Cforallt>0. (1.9)

    In particular, one has 0w(x,t)K0 for all (x,t)Ω×(0,), where

    K0:=max{||w0||L(Ω),K}. (1.10)

    Remark 1.1. For the special case F1(w)=F2(w)=w, h1(u)=u,h2(v)=v and f(w)=μw(1w), the results of Theorem 1.1 have been obtained in [45]. However, since α1 and α2 may be equal to zero in the hypotheses of this paper, the L2-spatiotemporal estimates of u and v cannot be directly obtained by using the method in [45]. By means of the mechanism "density-suppressed motility", we invoke some ideas used in [14] and apply the self-adjoint realization of Δ+δ with some δ>0 in L2(Ω) to establish the key L2-spatiotemporal estimates of u and v.

    The second main aim of this paper is to study the role of non-random dispersal and competition between two predators in the asymptotic properties of the nonnegative spatial homogeneous equilibria of System (1.8). For simplicity, we let F1(w)=F2(w)=w, h1(u)=θ1+α1u, h2(v)=θ2+α2v and f(w)=μw(1w), where θ1=θ2=θ>0 and α1,α2,μ>0; then, System system (1.8) can be simplified as

    {ut=Δ(d1(w)u)+γ1uwu(θ+α1u)β1uv,(x,t)Ω×(0,),vt=Δ(d2(w)v)+γ2vwv(θ+α2v)β2uv,(x,t)Ω×(0,),wt=DΔw(u+v)w+μw(1w),(x,t)Ω×(0,),uν=vν=wν=0,(x,t)Ω×(0,),(u,v,w)(x,0)=(u0,v0,w0)(x),xΩ. (1.11)

    Theorem 1.1 ensures that System (1.11) possesses a unique global bounded nonnegative classical solution (u,v,w) such that 0w(x,t)K0:=max{||w0||L(Ω),1} for all (x,t)Ω×(0,). Now we find some sufficient conditions of parameters so that (1.11) admits a positive constant solution (u,v,w) satisfying

    {γ1wθα1uβ1v=0,γ2wθα2vβ2u=0,uv+μμw=0, (1.12)

    i.e.,

    AX=B, (1.13)

    where

    A=[α1β1γ1β2α2γ211μ],X=[uvw],B=[θθμ].

    If the determinant Φ of the coefficient matrix A in (1.13) does not equal zero, it follows from Cramer's rule that

    u=ΦuΦ,v=ΦvΦ,w=ΦwΦ, (1.14)

    where

    {Φ=(γ1+β1μ)(α2β2)+(γ2+α2μ)(α1β1),Φu=α2μ(γ1θ)β1μ(γ2θ)+θ(γ1γ2),Φv=α1μ(γ2θ)β2μ(γ1θ)+θ(γ2γ1),Φw=(θ+β1μ)(α2β2)+(θ+α2μ)(α1β1). (1.15)

    When β1<α1 and β2<α2, it follows that Φ>0 and Φw>0, and thus we know w>0. Next, we shall discuss the sign of Φu and Φv. For convenience, we let

    β1:=l1μ(γ2θ) (1.16)

    and

    β2:=l2μ(γ1θ), (1.17)

    where

    l1:=α2μ(γ1θ)+θ(γ1γ2) (1.18)

    and

    l2:=α1μ(γ2θ)+θ(γ2γ1). (1.19)

    It is not difficult to see that u=ΦuΦ, v=ΦvΦ and w=ΦwΦ are positive, if γi>θ, i=1,2 and one of the following conditions holds:

    (H6) γ1<γ2, l1>0, β1<min{α1,β1} and β2<min{α2,β2};

    (H7) γ1>γ2, l2>0, β1<min{α1,β1} and β2<min{α2,β2};

    (H8) γ1=γ2 and max{β1,β2}<min{α1,α2}.

    Now, we give our main results on the asymptotic stability properties of the nonnegative spatial homogeneous equilibria of System (1.11) as follows.

    Theorem 1.2. Let ΩR2 be a smooth bounded domain and the parameters γi,αi,βi, (i=1,2), θ, μ and D be positive. Assume that d1(w) and d2(w) satisfy (H1), and that (u,v,w) is a global bounded classical solution of System (1.11). Suppose that γi>θ,i=1,2,

    (β1γ2+β2γ1)2<4γ1γ2α1α2 (1.20)

    and

    D>maxw[0,K0]w24w[u|d1(w)|2γ1d1(w)+v|d2(w)|2γ2d2(w)] (1.21)

    as well as one of the conditions (H6)–(H8) holds. Then for all initial data (u0,v0,w0) satisfying (H5), there exist positive constants C and λ such that

    ||u(,t)u||L(Ω)+||v(,t)v||L(Ω)+||w(,t)w||L(Ω)Ceλt (1.22)

    for all t>0, where (u,v,w) is given by (1.14).

    Remark 1.2. From a biological point of view, it is well known that the change of the predators comes from predation, competition and mortality in System (1.11). The parameters γi, βi, αi(i=1,2) and θ respectively stand for the predation rate, competition strength, density-dependent death and natural death rate of the predators, which play a collective role in studying the dynamical behavior of (1.11). More specifically, when γi>θ, it is called strong predation; otherwise, it is weak predation. Hence the results of Theorem 1.2 can tell us that if the predations of two predators are strong and the prey diffusion coefficient D is suitably large, all species can reach a coexistence state.

    Theorem 1.3. Let ΩR2 be a smooth bounded domain and the parameters γi,αi,βi, (i=1,2), θ, μ and D be positive. Assume that d1(w) and d2(w) satisfy (H1), and the (u,v,w) is a global bounded classical solution of System (1.11). Suppose that we have (1.20) and

    D>maxw[0,K0]¯uw2|d1(w)|24γ1¯wd1(w) (1.23)

    as well as one of the following conditions holds:

    (i) γ1>γ2>θ, l20, β1<min{α1,β1} and β2<α2;

    (ii) γ1>γ2>θ, l2>0, β1<min{α1,β1} and β2[β2,α2);

    (iii) γ1>θγ2,

    where

    ¯u=μ(γ1θ)α1μ+γ1and¯w=α1μ+θα1μ+γ1. (1.24)

    Then for all initial data (u0,v0,w0) satisfying (H5), one has

    ||u(,t)¯u||L(Ω)+||v(,t)||L(Ω)+||w(,t)¯w||L(Ω)0ast, (1.25)

    exponentially if θ>γ2¯wβ2¯u or algebraically if θ=γ2¯wβ2¯u.

    Theorem 1.4. Let ΩR2 be a smooth bounded domain and the parameters γi,αi,βi, (i=1,2), θ, μ and D be positive. Assume that d1(w) and d2(w) satisfy (H1), and the (u,v,w) is a global bounded solution of System (1.11). Suppose that we have (1.20) and

    D>maxw[0,K0]˜vw2|d2(w)|24γ2˜wd2(w) (1.26)

    as well as one of the following conditions holds:

    (i) γ2>γ1>θ, l10, β2<min{α2,β2} and β1<α1;

    (ii) γ2>γ1>θ, l1>0, β2<min{α2,β2} and β1[β1,α1);

    (iii) γ2>θγ1,

    where

    ˜v=μ(γ2θ)α2μ+γ2and˜w=α2μ+θα2μ+γ2. (1.27)

    Then for all initial data (u0,v0,w0) satisfying (H5), one has

    ||u(,t)||L(Ω)+||v(,t)˜v||L(Ω)+||w(,t)˜w||L(Ω)0ast, (1.28)

    exponentially if θ>γ1˜wβ1˜v or algebraically if θ=γ1˜wβ1˜v.

    Remark 1.3. It follows from Theorem 1.3 that if the predator u is superior over v in the competition and the prey diffusion coefficient D is suitably large, the semi-trivial equilibrium (¯u,0,¯w) is globally asymptotically stable. Similarly, we can obtain Theorem 1.4. Hence, we only give the conclusion of Theorem 1.4, without showing the details of the proof for brevity.

    Theorem 1.5. Let ΩR2 be a smooth bounded domain and the parameters γi,αi,βi, (i=1,2), θ, μ and D be positive. Assume that d1(w) and d2(w) satisfy (H1), and the (u,v,w) is a global bounded solution of System (1.11). Suppose that

    γiθ,i=1,2.

    Then for all initial data (u0,v0,w0) satisfying (H5), one has

    ||u(,t)||L(Ω)+||v(,t)||L(Ω)+||w(,t)1||L(Ω)0ast, (1.29)

    exponentially if γi<θ, i=1,2 or algebraically if γi=θ, i=1,2.

    Remark 1.4. It follows from Theorem 1.5 that if the capture rates of the two predators are low (i.e. γiθ, i=1,2), the prey-only steady state (0,0,1) is globally asymptotically stable regardless of the size of the prey diffusion coefficient D.

    Remark 1.5. Compared with the previous results of [33] without competitive terms, the results of Theorems 1.2–1.5 indicate that the competition terms play a crucial role in the global stability of the constant steady states in (1.11). Moreover, under the condition of density-suppressed motility, our global stability results of Theorems 1.2–1.5 can also generalize the ranges of parameters αi and βi(i=1,2) for two dimensional cases in [32]. However, since the heat-semigroup estimates of u and v are no longer applicable due to the appearance of density-suppressed motility, the global stability in L-norm is still open in the higher-dimensional problem.

    The rest of this paper is organized as follows. In Section 2, we first state the local existence of the classical solution to (1.8) and collect preliminary lemmas. In Section 3, we derive the global existence and boundedness of classical solutions for (1.8) and prove Theorem 1.1. Finally, we shall study the asymptotic stability of global bounded solutions for (1.11) and prove Theorems 1.2–1.5.

    In this section, we shall give the local existence and some preliminary lemmas. Firstly, we state the local existence of the classical solution to (1.8), as obtained by means of the abstract theory of quasilinear parabolic systems in [52].

    Lemma 2.1. Let D,γi,βi>0(i=1,2), ΩR2 be a smooth bounded domain and the hypotheses (H1)–(H5) hold. Then, there exists a Tmax(0,] such that System (1.8) possesses a unique classical solution (u,v,w)(C(¯Ω×[0,Tmax))C2,1(¯Ω×(0,Tmax)))3 satisfying

    u,v0and0wK0:=max{||w0||L(Ω),K}. (2.1)

    In addition, the following extensibility criterion holds, i.e. if Tmax<, then

    lim suptTmax(u(,t)L(Ω)+v(,t)L(Ω)+w(,t)W1,(Ω))=. (2.2)

    Proof. Let z=(u,v,w)T; then, System (1.8) can be rewritten as

    {zt=(P(z)z)+Q(z),(x,t)Ω×(0,),zν=0,(x,t)Ω×(0,),z(,0)=z0=(u0,v0,w0),xΩ, (2.3)

    where

    P(z)=(d1(w)0ud1(w)0d2(w)vd2(w)00D)andQ(z)=(γ1uF1(w)uh1(u)β1uvγ2vF2(w)vh2(v)β2uvuF1(w)vF2(w)+f(w)). (2.4)

    According to the conditions that D>0 and di(w)>0(i=1,2), the matrix P(z) is positively definite for the given initial data z0, which asserts that System (1.8) is normally parabolic. Thus it follows from Theorem 7.3 of [53] that there exists a Tmax(0,] such that System (1.8) admits a unique classical solution (u,v,w)(C0(¯Ω×[0,Tmax))C2,1(¯Ω×(0,Tmax)))3. The nonnegativity of (u,v,w) directly comes from the maximum principle [14,45]. It similarly follows from Lemma 2.2 of [13] that wK0:=max{||w0||L(Ω),K}. Since P(z) is an upper triangular matrix, we can deduce from Theorem 5.2 of [54] that the extensibility criterion given by (2.2) holds. The proof of Lemma 2.1 is complete.

    Lemma 2.2. Let ΩRn(n1) be a smooth bounded domain, D>0 and T(0,]. Assume that φ(x,t)C(¯Ω×[0,T))C2,1(¯Ω×(0,T)) satisfies

    {φt=DΔφφ+ψ,(x,t)Ω×(0,T),φν=0,(x,t)Ω×(0,T), (2.5)

    where ψL((0,T);Lp(Ω)) with p1. Then there exists a positive constant C such that

    ||φ(,t)||W1,q(Ω)Cforallt(0,T), (2.6)

    where

    q{[1,npnp),ifp<n,[1,),ifp=n,[1,],ifp>n. (2.7)

    Proof. This lemma directly comes from Lemma 1 of [55].

    Now, we give the following lemma ([56], Lemma 3.4) to derive some a priori estimates for w.

    Lemma 2.3. Let T>0,τ(0,T) and a,d>0, and assume that y:[0,T)[0,) is absolutely continuous. If there exists a nonnegative function hL1loc([0,T)) satisfying

    t+τth(s)dsdforallt[0,Tτ) (2.8)

    and

    y(t)+ay(t)h(t), (2.9)

    one has

    y(t)max{y(0)+d,daτ+2d}forallt[0,T). (2.10)

    Next, we give a basic property of the solution components u and v for System (1.8).

    Lemma 2.4. Let the assumptions of Lemma 2.1 hold. Then there exists a constant C>0 such that

    Ωu+vdxCforallt(0,Tmax) (2.11)

    and

    t+τtΩu2+v2dxdsCforallt(0,Tmaxτ), (2.12)

    where τ=min{1,12Tmax}.

    Proof. It follows from a direct computation for System (1.8) that

    ddtΩ(1γ1u+1γ2v+w)dx=Ωf(w)dx1γ1Ωuh1(u)dxβ1γ1Ωuvdx1γ2Ωvh2(v)dxβ2γ2ΩuvdxμΩwdx1γ1Ωu(θ1+α1u)dx1γ2Ωv(θ2+α2v)dx=μΩwdxθ1Ω1γ1udxθ2Ω1γ2vdxα1γ1Ωu2dxα2γ2Ωv2dx, (2.13)

    for all t(0,Tmax), where we have applied (H3), (H4), β1,β2>0 and (2.1).

    Let θ:=min{θ1,θ2}, it follows from (2.1) that

    ddtΩ(1γ1u+1γ2v+w)dx+α1γ1Ωu2dx+α2γ2Ωv2dxθΩ(1γ1u+1γ2v+w)dx+(μK0+θ)|Ω|, (2.14)

    which leads to (2.11) by Gronwall's inequality. If αi>0, i=1,2, then integrating (2.14) over (t,t+τ), we have (2.12) directly. If αi=0, i=1,2, we can also prove (2.12) by means of the idea used in [14]. For the readers' convenience, we give the sketch of the proof.

    Let A denote the self-adjoint realization of Δ+δ under homogeneous Neumann boundary conditions in L2(Ω), where δ(0,min{θ1d1(0),θ2d2(0)}). It follows from δ>0 that A has an order-preserving bounded inverse A1 on L2(Ω). Thus this allows us to obtain a positive constant c1 such that

    ||A1Ψ||L2(Ω)c1||Ψ||L2(Ω)for allΨL2(Ω) (2.15)

    and

    ||A12Ψ||2L2(Ω)=ΩΨA1Ψdxc1||Ψ||2L2(Ω)for allΨL2(Ω). (2.16)

    By a simple calculation in (1.8), we have

    (1γ1u+1γ2v+w)t=Δ(1γ1d1(w)u+1γ2d2(w)v+Dw)1γ1uh1(u)β1γ1uv1γ2vh2(v)β2γ2uv+f(w), (2.17)

    which can be written as

    (1γ1u+1γ2v+w)t+A(1γ1d1(w)u+1γ2d2(w)v+Dw)=δ(1γ1d1(w)u+1γ2d2(w)v+Dw)1γ1uh1(u)β1γ1uv1γ2vh2(v)β2γ2uv+f(w)=1γ1u(δd1(w)h1(u))+1γ2v(δd2(w)h2(v))+δDw+f(w)β1γ1uvβ2γ2uv1γ1u(δd1(0)θ1)+1γ2v(δd2(0)θ2)+(δD+μ)K0(δD+μ)K0:=c2, (2.18)

    where we have applied (H1), (H3), (H4), (2.1) and δ(0,min{θ1d1(0),θ2d2(0)}). Hence, by multiplying (2.18) by A1(1γ1u+1γ2v+w)0 and integrating it over Ω, we derive

    12ddtΩ|A12(1γ1u+1γ2v+w)|2dx+Ω(1γ1d1(w)u+1γ2d2(w)v+Dw)(1γ1u+1γ2v+w)dxc2ΩA1(1γ1u+1γ2v+w)dx. (2.19)

    According to the fact that 0<di(K0)di(w), i=1,2, due to (H1) and (2.1), and by letting c3:=min{d1(K0),d2(K0),D}>0, we deduce

    ddtΩ|A12(1γ1u+1γ2v+w)|2dx+2c3Ω(1γ1u+1γ2v+w)2dx2c2ΩA1(1γ1u+1γ2v+w)dx. (2.20)

    It follows from Hölder's and Young's inequality as well as (2.15) that

    2c2ΩA1(1γ1u+1γ2v+w)dx2c2|Ω|12(Ω|A1(1γ1u+1γ2v+w)|2dx)122c1c2|Ω|12(Ω(1γ1u+1γ2v+w)2dx)12c32Ω(1γ1u+1γ2v+w)2dx+2c21c22|Ω|c3. (2.21)

    According to (2.16), we have

    c32c1Ω|A12(1γ1u+1γ2v+w)|2c32Ω(1γ1u+1γ2v+w)2dx. (2.22)

    By combining (2.20)–(2.22), we derive

    ddtΩ|A12(1γ1u+1γ2v+w)|2dx+c32c1Ω|A12(1γ1u+1γ2v+w)|2+c3Ω(1γ1u+1γ2v+w)2dx2c21c22|Ω|c3. (2.23)

    By the ordinary differential equations (ODE) argument, there exists a c4>0 such that

    Ω|A12(1γ1u+1γ2v+w)|2dxc4, (2.24)

    which implies that

    t+τtΩ1γ21u2+1γ22v2dxdst+τtΩ(1γ1u+1γ2v)2dxdst+τtΩ(1γ1u+1γ2v+w)2dxdsc4c3+2c21c22|Ω|c23. (2.25)

    The proof of Lemma 2.4 is complete.

    Finally, we shall give the following key estimate of w, which plays a crucial role in the proof of Theorem 1.1.

    Lemma 2.5. Let the assumptions of Lemma 2.1 hold. Then there exists a constant C>0 such that

    Ω|w|2dxCforallt(0,Tmax) (2.26)

    and

    t+τtΩ|Δw|2dxdsCforallt(0,Tmaxτ), (2.27)

    where τ=min{1,12Tmax}.

    Proof. Multiplying the third equation of System (1.8) with 2Δw and integrating it by parts, we deduce from Young's inequality and (2.1) that

    ddtΩ|w|2dx=2DΩ|Δw|2dx+2Ω(uF1(w)+vF2(w))Δwdx2Ωf(w)ΔwdxDΩ|Δw|2dx+2DΩ(uF1(w)+vF2(w))2dx+2DΩf2(w)dxDΩ|Δw|2dx+4F21(K0)DΩu2dx+4F22(K0)DΩv2dx+2(μK0)2D|Ω|, (2.28)

    where we have used the hypotheses (H2) and (H4).

    It follows from wν=0, Young's inequality and (2.1) that

    Ω|w|2dx=ΩwΔwdxD2Ω|Δw|2dx+12DΩw2dxD2Ω|Δw|2dx+K202D|Ω|. (2.29)

    Let c5:=max{4F21(K0)D,4F22(K0)D}; we infer from (2.28) and (2.29) that

    ddtΩ|w|2dx+Ω|w|2dx+D2Ω|Δw|2dxc5Ωu2+v2dx+c6, (2.30)

    where c6:=2(μK0)2D|Ω|+K202D|Ω|. It follows from Lemma 2.3 and Lemma 2.4 that (2.26) holds. Then integrating (2.30) over (t,t+τ), we can deduce from (2.12) and (2.26) that (2.27) holds.

    In this section, we shall study the global existence and uniform boundedness of solutions for system (1.8) when n=2. To do this, we need the following lemmas.

    Lemma 3.1. Let the conditions of Theorem 1.1 hold. Then the solution (u,v,w) of system (1.8) satisfies

    Ωu2dx+Ωv2dxC (3.1)

    and

    ||w(,t)||W1,q(Ω)C (3.2)

    for all q[1,) and t(0,Tmax), where C>0 is a constant independent of t.

    Proof. Multiplying the first equation of System (1.8) by 2u and integrating by parts, we deduce from Young's and Hölder's inequalities that

    ddtΩu2dx=2Ωd1(w)|u|2dx2Ωd1(w)uuwdx+2γ1Ωu2F1(w)dx2Ωu2h1(u)dx2β1Ωu2vdxΩd1(w)|u|2dx+Ω|d1(w)|2d1(w)u2|w|2dx+2γ1F1(K0)Ωu2dxd1(K0)Ω|u|2dx+K1Ωu2|w|2dx+2γ1F1(K0)Ωu2dxd1(K0)Ω|u|2dx+K1(Ωu4dx)12(Ω|w|4dx)12+2γ1F1(K0)Ωu2dx, (3.3)

    where \mathcal{K}_{1}: = \max_{w\in[0, K_{0}]}\frac{|d'_{1}(w)|^{2}}{d_{1}(w)} and we have applied ( H_{1} )–( H_{3} ) and (2.1).

    By using the Gagliardo-Nirenberg inequality in two dimensions, there exists a C_{1} > 0 such that

    \begin{equation} \begin{split} \left(\int_{\Omega}u^{4}dx\right)^{\frac{1}{2}} = ||u||_{L^{4}(\Omega)}^{2}\leq C_{1}(||\nabla u||_{L^{2}(\Omega)}||u||_{L^{2}(\Omega)}+||u||_{L^{2}(\Omega)}^{2}). \end{split} \end{equation} (3.4)

    According to Lemma 2.5 of [19] when n = 2 , it follows from Lemma 2.5 that

    \begin{equation} \begin{split} \left(\int_{\Omega}|\nabla w|^{4}dx\right)^{\frac{1}{2}}&\leq C_{2}(||\Delta w||_{L^{2}(\Omega)}||\nabla w||_{L^{2}(\Omega)}+||\nabla w||_{L^{2}(\Omega)}^{2})\\ &\leq C_{3}(||\Delta w||_{L^{2}(\Omega)}+1), \end{split} \end{equation} (3.5)

    for all t\in(0, T_{\max}) , where C_{2}, C_{3} > 0 . Thus, by combining (3.4) with (3.5), we infer from Young's inequality that

    \begin{equation} \begin{split} \mathcal{K}_{1}\left(\int_{\Omega}u^{4}dx\right)^{\frac{1}{2}}\left(\int_{\Omega}|\nabla w|^{4}dx\right)^{\frac{1}{2}}&\leq d_{1}(K_{0})\int_{\Omega}|\nabla u|^{2}dx +C_{4}||u||_{L^{2}(\Omega)}^{2}||\Delta w||_{L^{2}(\Omega)}^{2}\\&\quad+C_{5}||u||_{L^{2}(\Omega)}^{2}, \end{split} \end{equation} (3.6)

    where C_{4}, C_{5} are positive constants. Thus it follows from (3.3) and (3.6) that

    \begin{equation} \begin{split} \frac{\text{d}}{\text{dt}}\int_{\Omega}u^{2}dx\leq C_{6}\int_{\Omega}u^{2}dx\left(\int_{\Omega}|\Delta w|^{2}dx+1\right), \end{split} \end{equation} (3.7)

    where C_{6}: = \max\{C_{4}, C_{5}+2\gamma_{1}F_{1}(K_{0})\} .

    By Lemma 2.4, we can find t_{0} = t_{0}(t)\in((t-\tau)_{+}, t) for any t\in(0, T_{\max}) such that there exists a C_{7} > 0 satisfying

    \begin{equation} \begin{split} \int_{\Omega}u^{2}(x,t_{0})dx\leq C_{7}, \end{split} \end{equation} (3.8)

    where \tau is defined in Lemma 2.4. It follows from Lemma 2.5 that there exists a C_{8} > 0 such that

    \begin{equation} \begin{split} \int_{t_{0}}^{t_{0}+\tau}\int_{\Omega}|\Delta w(x,t)|^{2}dxdt\leq C_{8}. \end{split} \end{equation} (3.9)

    Therefore, integrating (3.7) over (t_{0}, t) , we deduce from t_{0} < t < t_{0}+\tau\leq t_{0}+1 , (3.8) and (3.9) that

    \begin{equation} \begin{split} \int_{\Omega}u^{2}dx\leq \int_{\Omega}u^{2}(x,t_{0})dx e^{C_{6}\int_{t_{0}}^{t}\left(\int_{\Omega}|\Delta w|^{2}dx+1\right)ds}\leq C_{7}e^{C_{6}(C_{8}+1)} \end{split} \end{equation} (3.10)

    for all t\in(0, T_{\max}) .

    Similarly, we obtain

    \begin{equation} \begin{split} \int_{\Omega}v^{2}dx \leq C_{9} \quad\text{for all}\;\;t\in(0,T_{\max}). \end{split} \end{equation} (3.11)

    It follows from the third equation of System (1.8), we know that w solves the following problem

    \begin{eqnarray} \left\{ \begin{split}{} &w_t = D\Delta w-w+g(u,v,w),&(x,t)\in \Omega\times (0,T_{\max}),\\ &\frac{\partial w}{\partial\nu} = 0,&(x,t)\in \partial\Omega\times (0,T_{\max}), \end{split} \right. \end{eqnarray} (3.12)

    where g(u, v, w) = w-uF_{1}(w)-vF_{2}(w)+f(w) . According to ( H_{2} ), ( H_{3} ) and (2.1), we infer from (3.10) and (3.11) that

    \begin{equation} \begin{split} ||g(u,v,w)||_{L^{2}(\Omega)}\leq C_{10}(||u||_{L^{2}(\Omega)}+||v||_{L^{2}(\Omega)}+1)\leq C_{11} \end{split} \end{equation} (3.13)

    for all t\in(0, T_{\max}) . Hence, it follows from Lemma 2.2 in two dimensions that (3.2) holds. The proof of Lemma 3.1 is complete.

    Next, we shall prove the boundedness of w in W^{1\infty}(\Omega) .

    Lemma 3.2. Let the conditions of Theorem 1.1 hold. Then the solution component w of system (1.8) satisfies

    \begin{equation} \begin{split} ||w(\cdot,t)||_{W^{1,\infty}(\Omega)}\leq C \end{split} \end{equation} (3.14)

    for all t\in(0, T_{\max}) , where C > 0 is a constant independent of t .

    Proof. Multiplying the first equation of System (1.8) by u^{2} and integrating by parts, we deduce from Young's and Hölder's inequalities that

    \begin{equation} \begin{split} \frac{1}{3}\frac{\text{d}}{\text{dt}}\int_{\Omega}u^{3}dx& = -2\int_{\Omega}d_{1}(w)u|\nabla u|^{2}dx-2\int_{\Omega}d'_{1}(w)u^{2}\nabla u\cdot\nabla wdx\\ &\quad+\gamma_{1}\int_{\Omega}u^{3}F_{1}(w)dx-\int_{\Omega}u^{3}h_{1}(u)dx-\beta_{1}\int_{\Omega}u^{3}vdx\\ &\leq-\int_{\Omega}d_{1}(w)u|\nabla u|^{2}dx+\int_{\Omega}\frac{|d'_{1}(w)|^{2}}{d_{1}(w)}u^{3}|\nabla w|^{2}dx\\&\quad+\gamma_{1}F_{1}(K_{0})\int_{\Omega}u^{3}dx-\theta_{1}\int_{\Omega}u^{3}dx\\ &\leq -\frac{4d_{1}(K_{0})}{9}\int_{\Omega}|\nabla u^{\frac{3}{2}}|^{2}dx+\mathcal{K}_{1}\int_{\Omega}u^{3}|\nabla w|^{2}dx\\&\quad+\gamma_{1}F_{1}(K_{0})\int_{\Omega}u^{3}dx-\theta_{1}\int_{\Omega}u^{3}dx\\ &\leq -\frac{4d_{1}(K_{0})}{9}\int_{\Omega}|\nabla u^{\frac{3}{2}}|^{2}dx+\mathcal{K}_{1}\left(\int_{\Omega}u^{6}dx\right)^{\frac{1}{2}}\left(\int_{\Omega}|\nabla w|^{4}dx\right)^{\frac{1}{2}}\\&\quad+\gamma_{1}F_{1}(K_{0})\int_{\Omega}u^{3}dx-\theta_{1}\int_{\Omega}u^{3}dx, \end{split} \end{equation} (3.15)

    for all t\in(0, T_{\max}) , where \mathcal{K}_{1} is defined in the proof of Lemma 3.1 and we have applied ( H_{1} )–( H_{3} ) and (2.1).

    It follows from Lemma 3.1 that there exist positive constants C_{1} and C_{2} such that ||\nabla w||_{L^{4}(\Omega)}\leq C_{1} and ||u||_{L^{2}(\Omega)}\leq C_{2} for all t\in(0, T_{\max}) . Then by using the Gagliardo-Nirenberg inequality and Young's inequality, we can find positive constants C_{i} , i = 3, \cdot\cdot\cdot, 6 such that

    \begin{equation} \begin{split} \mathcal{K}_{1}\left(\int_{\Omega}u^{6}dx\right)^{\frac{1}{2}}\left(\int_{\Omega}|\nabla w|^{4}dx\right)^{\frac{1}{2}}&\leq\mathcal{K}_{1}C_{1}^{2}||u^{\frac{3}{2}}||_{L^{4}(\Omega)}^{2}\\ &\leq C_{3}\left(||\nabla u^{\frac{3}{2}}||_{L^{2}(\Omega)}^{\frac{4}{3}}\cdot||u^{\frac{3}{2}}||_{L^{\frac{4}{3}}(\Omega)}^{\frac{2}{3}}+||u^{\frac{3}{2}}||_{L^{\frac{4}{3}}(\Omega)}^{2}\right)\\ &\leq\frac{2d_{1}(K_{0})}{9}\int_{\Omega}|\nabla u^{\frac{3}{2}}|^{2}dx+C_{4} \end{split} \end{equation} (3.16)

    and

    \begin{equation} \begin{split} \gamma_{1}F_{1}(K_{0})\int_{\Omega}u^{3}dx&\leq\gamma_{1}F_{1}(K_{0})||u^{\frac{3}{2}}||_{L^{2}(\Omega)}^{2}\\ &\leq C_{5}\left(||\nabla u^{\frac{3}{2}}||_{L^{2}(\Omega)}^{\frac{2}{3}}\cdot||u^{\frac{3}{2}}||_{L^{\frac{4}{3}}(\Omega)}^{\frac{4}{3}}+||u^{\frac{3}{2}}||_{L^{\frac{4}{3}}(\Omega)}^{2}\right)\\ &\leq\frac{2d_{1}(K_{0})}{9}\int_{\Omega}|\nabla u^{\frac{3}{2}}|^{2}dx+C_{6} \end{split} \end{equation} (3.17)

    for all t\in(0, T_{\max}) .

    Combining (3.15)–(3.17), we have

    \begin{equation} \begin{split} \frac{\text{d}}{\text{dt}}\int_{\Omega}u^{3}dx+3\theta_{1}\int_{\Omega}u^{3}dx\leq C_{7}: = 3(C_{4}+C_{6}) \end{split} \end{equation} (3.18)

    for all t\in(0, T_{\max}) . By the ODE argument, we can derive

    \begin{equation} \begin{split} \int_{\Omega}u^{3}dx\leq \max\left\{\int_{\Omega}u_{0}^{3}dx,\frac{C_{7}}{3\theta_{1}}\right\} \quad \text{for all}\;\;t\in (0,T_{\max}). \end{split} \end{equation} (3.19)

    Similarly, we also derive the boundedness of ||v||_{L^{3}(\Omega)} . Then it follows from Lemma 2.2 in two dimensions that (3.14) holds.

    By means of the boundedness of ||w(\cdot, t)||_{W^{1, \infty}(\Omega)} , it follows from the Moser iteration of [45] that we can obtain the boundedness of ||u(\cdot, t)||_{L^{\infty}(\Omega)} and ||v(\cdot, t)||_{L^{\infty}(\Omega)} for all t\in(0, T_{\max}) .

    Lemma 3.3. Let the conditions of Theorem 1.1 hold. Then the solution component (u, v) of system (1.8) satisfies

    \begin{equation} \begin{split} ||u(\cdot,t)||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}\leq C \end{split} \end{equation} (3.20)

    for all t\in(0, T_{\max}) , where C > 0 is a constant independent of t .

    Proof. Multiplying the first equation of System (1.8) by u^{p-1} with p\geq2 and integrating by parts, we deduce from Young's inequality that

    \begin{equation} \begin{split} \frac{1}{p}\frac{\text{d}}{\text{dt}}\int_{\Omega}u^{p}dx& = -(p-1)\int_{\Omega}d_{1}(w)u^{p-2}|\nabla u|^{2}dx-(p-1)\int_{\Omega}d'_{1}(w)u^{p-1}\nabla u\cdot\nabla wdx\\ &\quad+\gamma_{1}\int_{\Omega}u^{p}F_{1}(w)dx-\int_{\Omega}u^{p}h_{1}(u)dx-\beta_{1}\int_{\Omega}u^{p}vdx\\ &\leq-\frac{p-1}{2}\int_{\Omega}d_{1}(w)u^{p-2}|\nabla u|^{2}dx+\frac{p-1}{2}\int_{\Omega}\frac{|d'_{1}(w)|^{2}}{d_{1}(w)}u^{p}|\nabla w|^{2}dx\\&\quad+\gamma_{1}F_{1}(K_{0})\int_{\Omega}u^{p}dx\\ &\leq -\frac{p-1}{2}d_{1}(K_{0})\int_{\Omega}u^{p-2}|\nabla u|^{2}dx+\frac{p-1}{2}\mathcal{K}_{1}\int_{\Omega}u^{p}|\nabla w|^{2}dx\\&\quad+\gamma_{1}F_{1}(K_{0})\int_{\Omega}u^{p}dx \end{split} \end{equation} (3.21)

    for all t\in(0, T_{\max}) , where \mathcal{K}_{1} is defined in the proof of Lemma 3.1 and we have applied ( H_{1} )–( H_{3} ) and (2.1).

    It follows from Lemma 3.2 that there exists a C_{1} > 0 such that ||\nabla w(\cdot, t)||_{L^{\infty}(\Omega)}\leq C_{1} for all t\in (0, T_{\max}) . Hence, we deduce from (3.21) that

    \begin{equation} \begin{split} &\frac{\text{d}}{\text{dt}}\int_{\Omega}u^{p}dx+\frac{p(p-1)d_{1}(K_{0})}{2}\int_{\Omega}u^{p-2}|\nabla u|^{2}dx+p(p-1)\int_{\Omega}u^{p}dx\\ &\qquad\leq C_{2}p(p-1)\int_{\Omega}u^{p}dx, \end{split} \end{equation} (3.22)

    for all t\in(0, T_{\max}) , where C_{2}: = \frac{\mathcal{K}_{1}C_{1}^{2}}{2}+\gamma_{1}F_{1}(K_{0})+1 is independent of p . The rest can be handled exactly as the Moser iteration in Lemma 2.7 of [45] to derive the boundedness of ||u(\cdot, t)||_{L^{\infty}(\Omega)} for all t\in(0, T_{\max}) . Similarly, we can obtain the boundedness of ||v(\cdot, t)||_{L^{\infty}(\Omega)} for all t\in(0, T_{\max}) . The proof of Lemma 3.3 is complete.

    Proof of Theorem 1.1. Theorem 1.1 is a direct consequence of Lemma 2.1, Lemma 3.2 and Lemma 3.3.

    In this section, we shall study the asymptotic stability of global bounded solutions for System (1.11) by constructing energy functionals used in [13,57]. To do this, we first give some regularity results of the solution (u, v, w) for System (1.11).

    Lemma 4.1. Let (u, v, w) be a global bounded classical solution for (1.11) ensured in Theorem 1.1. Then there exist \sigma\in(0, 1) and C > 0 such that

    \begin{equation} \begin{split} ||u||_{C^{\sigma,\frac{\sigma}{2}}(\overline{\Omega}\times [t,t+1])}+||v||_{C^{\sigma,\frac{\sigma}{2}}(\overline{\Omega}\times [t,t+1])}+||w||_{C^{2+\sigma,1+\frac{\sigma}{2}}(\overline{\Omega}\times [t,t+1])}\leq C\quad{for \;all}\;\;t > 1. \end{split} \end{equation} (4.1)

    Proof. This lemma can be verified by a similar argument in Lemma 4.1 of [14], so we omit the details here for brevity.

    Lemma 4.2. Let (u, v, w) be a global bounded classical solution for (1.11) ensured in Theorem 1.1. Then there exists a C > 0 such that

    \begin{equation} \begin{split} ||u(\cdot,t)||_{W^{1,4}(\Omega)}+||v(\cdot,t)||_{W^{1,4}(\Omega)}\leq C\quad{for \;all}\;\;t > 0. \end{split} \end{equation} (4.2)

    Proof. This lemma can be verified by a similar argument in Lemma 3.6 of [45], so we omit the details here for brevity.

    In order to prove the asymptotic stabilization of global bounded solutions for system (1.11), we provide the following lemma, which is proved in [57].

    Lemma 4.3. Let \phi:(1, \infty)\rightarrow [0, \infty) be uniformly continuous such that \int_{1}^{\infty}\phi(t)dt < \infty . Then

    \begin{equation} \phi(t)\rightarrow 0\;\;{as}\;\;t\rightarrow \infty. \end{equation} (4.3)

    In this subsection, we are devoted to studying the stabilization of the coexistence steady state (u^{*}, v^{*}, w^{*}) for some parameters cases. Let us introduce the following functionals

    \begin{aligned} \mathcal{E}_{1}(t) = &\frac{1}{\gamma_{1}}\int_{\Omega}\left(u-u^{*}-u^{*} \ln \frac{u }{u^{*}}\right)dx+\frac{1}{\gamma_{2}}\int_{\Omega}\left(v-v^{*}-v^{*} \ln \frac{v }{v^{*}}\right)dx \\ &+\int_{\Omega}\left(w-w^{*}-w^{*} \ln \frac{w }{w^{*}}\right)dx,\\ \end{aligned}

    and

    \begin{aligned} \mathcal{F}_{1}(t) = &\int_{\Omega}\left(u -u^{*}\right)^{2}dx+\int_{\Omega}\left(v -v^{*}\right)^{2}dx+ \int_{\Omega}\left(w-w^{*}\right)^{2}dx\\ &+\int_\Omega {{{\left| {\frac{{\nabla u}}{u}} \right|}^2}}dx+\int_\Omega{{{\left| {\frac{{\nabla v}}{v}} \right|}^2}}dx+\int_\Omega {{{\left| {\nabla w} \right|}^2}}dx, \end{aligned}

    where (u^{*}, v^{*}, w^{*}) is given by (1.14).

    Lemma 4.4. Let the conditions of Theorem 1.2 hold. Then there exists a positive constant \varepsilon_{1} independent of t such that

    \begin{equation} \begin{aligned} \mathcal{E}_{1}(t) \geq 0 \end{aligned} \end{equation} (4.4)

    and

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{1}(t) \leq-\varepsilon_{1} \mathcal{F}_{1}(t) \;\;\;{for\; all\; t > 0 } . \end{aligned} \end{equation} (4.5)

    Proof. Let

    \begin{array}{l} I_{1}(t): = \frac{1}{\gamma_{1}}\int_{\Omega}\left(u-u^{*}-u^{*} \ln \frac{u}{u^{*}}\right)dx, \\ I_{2}(t): = \frac{1}{\gamma_{2}}\int_{\Omega}\left(v-v^{*}-v^{*} \ln \frac{v}{v^{*}}\right)dx, \\ I_{3}(t): = \int_{\Omega}\left(w-w^{*}-w^{*} \ln \frac{w }{w^{*}}\right)dx, \end{array}

    then \mathcal{E}_{1}(t) can be rewritten as

    \mathcal{E}_{1}(t) = I_{1}(t)+I_{2}(t)+I_{3}(t)\quad\text{for all}\;\;t > 0.

    Step 1: We shall prove the nonnegativity of \mathcal{E}_{1}(t) for all t > 0 . Let H(\xi): = \xi-u^{*}\ln \xi for \xi > 0 ; it follows from Taylor's formula for all x \in \Omega and each t > 0 that there exists a \tau = \tau(x, t) \in(0, 1) such that

    \begin{aligned} u-u^{*}-u^{*} \ln \frac{u}{u^{*}}& = H(u)-H\left(u^{*}\right) \\ & = H^{\prime}\left(u^{*}\right) \cdot\left(u-u^{*}\right)+\frac{1}{2} H^{\prime \prime}\left(\tau u+(1-\tau) u^{*}\right) \cdot\left(u-u^{*}\right)^{2} \\ & = \frac{u^{*}}{2\left(\tau u+(1-\tau) u^{*}\right)^{2}}\left(u-u^{*}\right)^{2} \geq 0. \end{aligned}

    Hence, we immediately derive that I_{1}(t) = \int_{\Omega}\left(H(u)-H\left(u^{*}\right)\right)dx \geq 0 . Similarly, we know that I_{2}(t) \geq 0 and I_{3}(t) \geq 0 for all t > 0 . Thus, we know that (4.4) holds.

    Step 2: Now, we further prove (4.5). By a series of simple calculations, we get

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} I_{1}(t)& = \frac{1}{\gamma_{1}}\int_{\Omega}\frac{u-u^{*}}{u}u_{t}dx\\ & = \frac{1}{\gamma_{1}}\int_{\Omega}\frac{u-u^{*}}{u}(\Delta (d_{1}(w)u)+\gamma_{1}uw-u(\theta+\alpha_{1} u)-\beta_{1}uv)dx\\ & = -\frac{u^{*}}{\gamma_{1}}\int_{\Omega}\frac{d_{1}(w)|\nabla u|^{2}}{u^{2}}dx-\frac{u^{*}}{\gamma_{1}}\int_{\Omega}\frac{d'_{1}(w)\nabla u\cdot\nabla w}{u}dx\\&\quad +\frac{1}{\gamma_{1}}\int_{\Omega}(u-u^{*})(\gamma_{1}w-\theta-\alpha_{1}u-\beta_{1}v)dx\\ & = -\frac{u^{*}}{\gamma_{1}}\int_{\Omega}\frac{d_{1}(w)|\nabla u|^{2}}{u^{2}}dx-\frac{u^{*}}{\gamma_{1}}\int_{\Omega}\frac{d'_{1}(w)\nabla u\cdot\nabla w}{u}dx\\&\quad +\int_{\Omega}(u-u^{*})(w-w^{*})dx-\frac{\alpha_{1}}{\gamma_{1}}\int_{\Omega}(u-u^{*})^{2}dx\\&\quad-\frac{\beta_{1}}{\gamma_{1}}\int_{\Omega}(u-u^{*})(v-v^{*})dx, \end{aligned} \end{equation} (4.6)

    where we have used the fact that \theta = \gamma_{1}w^{*}-\alpha_{1}u^{*}-\beta_{1}v^{*} .

    Similarly, it follows from the identities \theta = \gamma_{2}w^{*}-\beta_{2}u^{*}-\alpha_{2}v^{*} and \mu = u^{*}+v^{*}+\mu w^{*} that

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} I_{2}(t) & = -\frac{v^{*}}{\gamma_{2}}\int_{\Omega}\frac{d_{2}(w)|\nabla v|^{2}}{v^{2}}dx-\frac{v^{*}}{\gamma_{2}}\int_{\Omega}\frac{d'_{2}(w)\nabla v\cdot\nabla w}{v}dx\\&\quad +\int_{\Omega}(v-v^{*})(w-w^{*})dx-\frac{\alpha_{2}}{\gamma_{2}}\int_{\Omega}(v-v^{*})^{2}dx\\&\quad-\frac{\beta_{2}}{\gamma_{2}}\int_{\Omega}(v-v^{*})(u-u^{*})dx, \end{aligned} \end{equation} (4.7)

    and

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} I_{3}(t) & = -Dw^{*}\int_{\Omega}\frac{|\nabla w|^{2}}{w^{2}}dx- \int_{\Omega}(w-w^{*})(u-u^{*})dx-\int_{\Omega}(w-w^{*})(v-v^{*})dx\\&\quad-\mu\int_{\Omega}(w-w^{*})^{2}dx. \end{aligned} \end{equation} (4.8)

    Hence, by combining (4.6)–(4.8), we derive

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}}\mathcal{E}_{1}(t)& = -\frac{\alpha_{1}}{\gamma_{1}}\int_{\Omega}(u-u^{*})^{2}dx -\frac{\alpha_{2}}{\gamma_{2}}\int_{\Omega}(v-v^{*})^{2}dx-\mu\int_{\Omega}(w-w^{*})^{2}dx\\&\quad -\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)\int_{\Omega}(u-u^{*})(v-v^{*})dx-\frac{u^{*}}{\gamma_{1}}\int_{\Omega}\frac{d_{1}(w)|\nabla u|^{2}}{u^{2}}dx\\&\quad-\frac{v^{*}}{\gamma_{2}}\int_{\Omega}\frac{d_{2}(w)|\nabla v|^{2}}{v^{2}}dx-Dw^{*}\int_{\Omega}\frac{|\nabla w|^{2}}{w^{2}}dx-\frac{u^{*}}{\gamma_{1}}\int_{\Omega}\frac{d'_{1}(w)\nabla u\cdot\nabla w}{u}dx\\ &\quad-\frac{v^{*}}{\gamma_{2}}\int_{\Omega}\frac{d'_{2}(w)\nabla v\cdot\nabla w}{v}dx\\ &: = -\int_{\Omega}X_{1}A_{1}X_{1}^{T}dx-\int_{\Omega}Y_{1}B_{1}Y_{1}^{T}dx, \end{aligned} \end{equation} (4.9)

    where X_{1} = (u-u^{*}, v-v^{*}, w-w^{*}) and Y_{1} = \left(\frac{\nabla u}{u}, \frac{\nabla v}{v}, \nabla w\right) , as well as

    \begin{equation} A_{1} = \left( \begin{array}{ccc} \frac{\alpha_{1}}{\gamma_{1}} &\frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right) &0\\ \frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right) &\frac{\alpha_{2}}{\gamma_{2}} &0\\ 0&0&\mu \end{array} \right), B_{1} = \left( \begin{array}{ccc} \frac{u^{*}d_{1}(w)}{\gamma_{1}} &0 &\frac{u^{*}d'_{1}(w)}{2\gamma_{1}}\\ 0 &\frac{v^{*}d_{2}(w)}{\gamma_{2}} &\frac{v^{*}d'_{2}(w)}{2\gamma_{2}}\\ \frac{u^{*}d'_{1}(w)}{2\gamma_{1}}&\frac{v^{*}d'_{2}(w)}{2\gamma_{2}}&\frac{Dw^{*}}{w^{2}} \end{array} \right). \end{equation} (4.10)

    It follows from (1.20) that

    \begin{equation} \begin{split} \left|\frac{\alpha_{1}}{\gamma_{1}}\right| > 0 \quad\text{and}\quad\left|\begin{matrix} \frac{\alpha_{1}}{\gamma_{1}} & \frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)\\ \frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right) & \frac{\alpha_{2}}{\gamma_{2}} \end{matrix}\right| = \frac{\alpha_{1}\alpha_{2}}{\gamma_{1}\gamma_{2}}-\frac{1}{4}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)^{2} > 0 \end{split} \end{equation} (4.11)

    as well as

    \begin{equation} |A_{1}| = \mu\left( \frac{\alpha_{1}\alpha_{2}}{\gamma_{1}\gamma_{2}}-\frac{1}{4}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)^{2}\right) > 0, \end{equation} (4.12)

    which implies that the matrix A_{1} is positive definite as according to Sylvester's criterion. Similarly, we deduce from (1.21) that

    \begin{equation} \begin{split} \left|\frac{u^{*}d_{1}(w)}{\gamma_{1}}\right| > 0 \quad\text{and}\quad\left|\begin{matrix} \frac{u^{*}d_{1}(w)}{\gamma_{1}} & 0\\ 0 & \frac{v^{*}d_{2}(w)}{\gamma_{2}} \end{matrix}\right| = \frac{u^{*}v^{*}d_{1}(w)d_{2}(w)}{\gamma_{1}\gamma_{2}} > 0 \end{split} \end{equation} (4.13)

    as well as

    \begin{equation} \begin{split} |B_{1}| = \frac{u^{*}v^{*}w^{*}d_{1}(w)d_{2}(w)}{\gamma_{1}\gamma_{2}w^{2}}\left(D-\frac{u^{*}w^{2}|d'_{1}(w)|^{2}}{4\gamma_{1}w^{*}d_{1}(w)} -\frac{v^{*}w^{2}|d'_{2}(w)|^{2}}{4\gamma_{2}w^{*}d_{2}(w)}\right) > 0, \end{split} \end{equation} (4.14)

    which implies that the matrix B_{1} is positive definite. Thus there exist positive constants \kappa_{1} and \kappa_{2} such that

    \begin{equation} \begin{split} X_{1}A_{1}X_{1}^{T}\geq \kappa_{1}|X_{1}|^{2}\quad\text{and}\quad Y_{1}B_{1}Y_{1}^{T}\geq \kappa_{2}|Y_{1}|^{2} \end{split} \end{equation} (4.15)

    for all x\in \Omega and t > 0 . Let \varepsilon_{1}: = \min\{\kappa_{1}, \kappa_{2}\} , we have

    \begin{equation} \begin{split} \frac{\text{d}}{\text{dt}}\mathcal{E}_{1}(t)\leq-\varepsilon_{1}\int_{\Omega}|X_{1}|^{2}+|Y_{1}|^{2}dx\quad\text{for all}\;\;t > 0, \end{split} \end{equation} (4.16)

    which implies that (4.5) holds. The proof of Lemma 4.4 is complete.

    With the aid of Lemma 4.4, we shall give the following large time behavior of global solutions for system (1.11).

    Lemma 4.5. Let the assumptions of Theorem 1.2 hold. Then the global bounded solution of (1.11) converges to the coexistence steady state (u^{*}, v^{*}, w^{*}) given by (1.14), i.e.,

    \begin{equation} ||u(\cdot,t)-u^{*}||_{L^{\infty}(\Omega)}+||v(\cdot,t)-v^{*}||_{L^{\infty}(\Omega)}+||w(\cdot,t)-w^{*}||_{L^{\infty}(\Omega)}\rightarrow 0 \end{equation} (4.17)

    as t\rightarrow \infty .

    Proof. It follows from Lemma 4.4 and integration over (1, \infty) that

    \int_{1}^{\infty} \mathcal{F}_{1}(t) d t \leq \frac{\mathcal{E}_{1}(1)}{\varepsilon_{1}} < \infty.

    According to Theorem 1.1 and Lemma 4.1, the bounded solution u, v and w are Hölder continuous in \bar{\Omega} \times[t, t+1] with respect to t > 1 . Thus we conclude that \mathcal{F}_{1}(t) is uniformly continuous in (1, \infty) . Thus we infer from Lemma 4.3 that

    \begin{equation} \begin{aligned} \int_{\Omega}\left(u-u^{*}\right)^{2}dx+\int_{\Omega}\left(v-v^{*}\right)^{2}dx+\int_{\Omega}\left(w-w^{*}\right)^{2}dx\rightarrow 0 \end{aligned} \end{equation} (4.18)

    as t \rightarrow \infty . By the Gagliardo-Nirenberg inequality in two dimensions, there exists a C_{1} > 0 such that

    \left\|u-u^{*}\right\|_{L^{\infty}(\Omega)} \leq C_{1}\left\|u-u^{*}\right\|_{W^{1, 4}(\Omega)}^{\frac{2}{3}}\left\|u-u^{*}\right\|_{L^{2}(\Omega)}^{\frac{1}{3}}.

    Moreover, it follows from Lemma 4.2 that u(\cdot, t)-u^{*} is bounded in W^{1, 4}(\Omega) ; thus, we conclude from (4.18) that u(\cdot, t) \rightarrow u^{*} in L^{\infty}(\Omega) as t \rightarrow \infty . By the similar arguments for v and w , we derive (4.17). The proof of Lemma 4.5 is complete.

    Now, we give the convergence rate of the coexistence state (u^{*}, v^{*}, w^{*}) for System (1.11).

    Lemma 4.6. Let the assumptions of Theorem 1.2 hold; the global bounded solution (u, v, w) of (1.11) exponentially converges to the coexistence state (u^{*}, v^{*}, w^{*}) , i.e. there exist C > 0 and \lambda > 0 such that

    \begin{equation} ||u(\cdot,t)-u^{*}||_{L^{\infty}(\Omega)}+||v(\cdot,t)-v^{*}||_{L^{\infty}(\Omega)}+||w(\cdot,t)-w^{*}||_{L^{\infty}(\Omega)}\leq Ce^{-\lambda t} \end{equation} (4.19)

    for all t > T_{1} , where T_{1} > 0 is some fixed time.

    Proof. It follows from Lemma 4.5 that ||u-u^{*}||_{L^{\infty}(\Omega)}\rightarrow 0 as t\rightarrow \infty . Therefore, we apply L'Hôpital's rule to obtain

    \begin{equation} \lim\limits_{u\rightarrow u^{*}}\frac{u-u^{*}-u^{*}\ln\frac{u}{u^{*}}}{(u-u^{*})^{2}} = \frac{1}{2u^{*}}, \end{equation} (4.20)

    which implies that there exists a t_{1} > 0 such that

    \begin{equation} \begin{split} \frac{1}{4u^{*}}\int_{\Omega}(u-u^{*})^{2}dx\leq\int_{\Omega}\left(u-u^{*}-u^{*}\ln\frac{u}{u^{*}}\right)dx\leq\frac{3}{4u^{*}}\int_{\Omega}(u-u^{*})^{2}dx \end{split} \end{equation} (4.21)

    for all t > t_{1} . Similarly, we can find t_{2} > 0 satisfying

    \begin{equation} \begin{split} \frac{1}{4v^{*}}\int_{\Omega}(v-v^{*})^{2}dx\leq\int_{\Omega}\left(v-v^{*}-v^{*}\ln\frac{v}{v^{*}}\right)dx\leq\frac{3}{4v^{*}}\int_{\Omega}(v-v^{*})^{2}dx \end{split} \end{equation} (4.22)

    and

    \begin{equation} \begin{split} \frac{1}{4w^{*}}\int_{\Omega}(w-w^{*})^{2}dx\leq\int_{\Omega}\left(w-w^{*}-w^{*}\ln\frac{w}{w^{*}}\right)dx\leq\frac{3}{4w^{*}}\int_{\Omega}(w-w^{*})^{2}dx \end{split} \end{equation} (4.23)

    for all t > t_{2} . Let T_{1}: = \max\{t_{1}, t_{2}\} ; by means of the definitions of \mathcal{E}_{1}(t) and \mathcal{F}_{1}(t) , it follows from the second inequalities in (4.21)–(4.23) that there exists a C_{1} > 0 such that

    \begin{equation} C_{1}\mathcal{E}_{1}(t)\leq \mathcal{F}_{1}(t)\;\;\text{for all}\; \;t > T_{1}. \end{equation} (4.24)

    By Lemma 4.4, we derive

    \begin{equation} \mathcal{E}_{1}'(t)\leq -\varepsilon_{1} \mathcal{F}_{1}(t)\leq-\varepsilon_{1} C_{1}\mathcal{E}_{1}(t)\quad\text{for all}\;\;t > T_{1}, \end{equation} (4.25)

    which implies that there exist C_{2} > 0 and C_{3} > 0 such that

    \begin{equation} \mathcal{E}_{1}(t)\leq C_{2}e^{-C_{3}(t-T_{1})}\quad\text{for all}\;\;t > T_{1}. \end{equation} (4.26)

    Thus we deduce from the first inequalities in (4.21)–(4.23) that there exists a C_{4} > 0 such that

    \begin{equation} \begin{split} &\int_{\Omega}(u(x,t)-u^{*})^{2}dx+\int_{\Omega}(v(x,t)-v^{*})^{2}dx+\int_{\Omega}(w(x,t)-w^{*})^{2}dx\\ &\quad\leq C_{4}\mathcal{E}_{1}(t)\leq C_{2}C_{4}e^{-C_{3}(t-T_{1})}\;\;\text{for all}\;\;t > T_{1}. \end{split} \end{equation} (4.27)

    It follows from the Gagliardo-Nirenberg inequality in two dimensions, Lemma 4.2 and Lemma 3.2 that there exist positive constants C_{5} and C_{6} such that

    \begin{equation} \begin{split} &||u-u^{*}||_{L^{\infty}(\Omega)}+||v-v^{*}||_{L^{\infty}(\Omega)}+||w-w^{*}||_{L^{\infty}(\Omega)}\\ &\leq C_{5}\bigg(||u-u^{*}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||u-u^{*}||_{L^{2}(\Omega)}^{\frac{1}{3}} +||v-v^{*}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||v-v^{*}||_{L^{2}(\Omega)}^{\frac{1}{3}}\\& \quad+||w-w^{*}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||w-w^{*}||_{L^{2}(\Omega)}^{\frac{1}{3}}\bigg)\\ &\leq C_{6}\left(\int_{\Omega}(u-u^{*})^{2}dx+\int_{\Omega}(v-v^{*})^{2}dx+\int_{\Omega}(w-w^{*})^{2}dx\right)^{\frac{1}{6}}\\ &\leq C_{6}(C_{2}C_{4})^{\frac{1}{6}}e^{\frac{-C_{3}(t-T_{1})}{6}} \end{split} \end{equation} (4.28)

    for all t > T_{1} . The proof of Lemma 4.6 is complete.

    Proof of Theorem 1.2. The statement of Theorem 1.2 is a straightforward consequence of Lemma 4.6.

    In this subsection, we shall study the stabilization of the semi-trivial steady state (\overline{u}, 0, \overline{w}) or (0, \widetilde{v}, \widetilde{w}) for some parameters cases. Since the methods of the proofs of Theorem 1.3 and Theorem 1.4 are very similar, we only give the proof of Theorem 1.3 for brevity. To do this, let us introduce the following functionals

    \begin{aligned} \mathcal{E}_{2}(t) = &\frac{1}{\gamma_{1}}\int_{\Omega}\left(u-\overline{u}-\overline{u}\ln \frac{u }{\overline{u}}\right)dx+\frac{1}{\gamma_{2}}\int_{\Omega}vdx \\ &+\int_{\Omega}\left(w-\overline{w}-\overline{w}\ln \frac{w}{\overline{w}}\right)dx,\\ \end{aligned}

    and

    \begin{aligned} \mathcal{F}_{2}(t) = &\int_{\Omega}\left(u -\overline{u}\right)^{2}dx+\int_{\Omega}v^{2}dx+ \int_{\Omega}\left(w-\overline{w}\right)^{2}dx\\ &+\int_\Omega {{{\left| {\frac{{\nabla u}}{u}} \right|}^2}}dx+\int_\Omega {{{\left| {\nabla w} \right|}^2}}dx, \end{aligned}

    where \overline{u} = \frac{\mu(\gamma_{1}-\theta)}{\alpha_{1}\mu+\gamma_{1}}\; \; \text{and}\; \; \overline{w} = \frac{\alpha_{1}\mu+\theta}{\alpha_{1}\mu+\gamma_{1}} .

    Lemma 4.7. Let the conditions of Theorem 1.3 hold. Then there exists a positive constant \varepsilon_{2} independent of t such that

    \begin{equation} \begin{aligned} \mathcal{E}_{2}(t) \geq 0 \end{aligned} \end{equation} (4.29)

    and

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{2}(t) \leq-\varepsilon_{2} \mathcal{F}_{2}(t)-\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u})\int_{\Omega}vdx \;\;\;{for\; all\; t > 0 }. \end{aligned} \end{equation} (4.30)

    Proof. Let

    \begin{array}{l} J_{1}(t): = \frac{1}{\gamma_{1}}\int_{\Omega}\left(u-\overline{u}- \overline{u}\ln \frac{u}{\overline{u}}\right)dx, \\ J_{2}(t): = \frac{1}{\gamma_{2}}\int_{\Omega}vdx, \\ J_{3}(t): = \int_{\Omega}\left(w-\overline{w}-\overline{w}\ln \frac{w }{\overline{w}}\right)dx, \end{array}

    then \mathcal{E}_{2}(t) can be represented as

    \mathcal{E}_{2}(t) = J_{1}(t)+J_{2}(t)+J_{3}(t)\quad\text{for all}\;\;t > 0.

    Firstly, we can prove the nonnegativity of \mathcal{E}_{2}(t) for all t > 0 by the similar arguments used in Step 1 in Lemma 4.4. For brevity, we omit the details here. Now, we shall prove (4.30). By a series of simple calculations, we get

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} J_{1}(t) & = -\frac{\overline{u}}{\gamma_{1}}\int_{\Omega}\frac{d_{1}(w)|\nabla u|^{2}}{u^{2}}dx-\frac{\overline{u}}{\gamma_{1}}\int_{\Omega}\frac{d'_{1}(w)\nabla u\cdot\nabla w}{u}dx\\&\quad +\int_{\Omega}(u-\overline{u})(w-\overline{w})dx-\frac{\alpha_{1}}{\gamma_{1}}\int_{\Omega}(u-\overline{u})^{2}dx\\&\quad -\frac{\beta_{1}}{\gamma_{1}}\int_{\Omega}(u-\overline{u})vdx, \end{aligned} \end{equation} (4.31)

    where we have used the fact that \theta = \gamma_{1}\overline{w}-\alpha_{1}\overline{u} .

    Similarly, we can derive

    \begin{equation} \begin{aligned} {\frac{\text{d}}{\text{dt}}} J_{2}(t) & = -\frac{\alpha_{2}}{\gamma_{2}}\int_{\Omega}v^{2}dx-\frac{\beta_{2}}{\gamma_{2}}\int_{\Omega}v(u-\overline{u})dx\\&\quad +\int_{\Omega}v(w-\overline{w})dx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u})\int_{\Omega}vdx, \end{aligned} \end{equation} (4.32)

    and

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} J_{3}(t) & = -D\overline{w}\int_{\Omega}\frac{|\nabla w|^{2}}{w^{2}}dx- \int_{\Omega}(w-\overline{w})(u-\overline{u})dx-\int_{\Omega}(w-\overline{w})vdx\\&\quad-\mu\int_{\Omega}(w-\overline{w})^{2}dx, \end{aligned} \end{equation} (4.33)

    where we have used the fact that \mu = \overline{u}+\mu \overline{w} . Thus it follows from (4.31)–(4.33) that

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{2}(t)& = -\frac{\alpha_{1}}{\gamma_{1}}\int_{\Omega}(u-\overline{u})^{2}dx -\frac{\alpha_{2}}{\gamma_{2}}\int_{\Omega}v^{2}dx-\mu\int_{\Omega}(w-\overline{w})^{2}dx\\&\quad -\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)\int_{\Omega}(u-\overline{u})vdx-\frac{\overline{u}}{\gamma_{1}}\int_{\Omega}\frac{d_{1}(w)|\nabla u|^{2}}{u^{2}}dx\\&\quad-D\overline{w}\int_{\Omega}\frac{|\nabla w|^{2}}{w^{2}}dx-\frac{\overline{u}}{\gamma_{1}}\int_{\Omega}\frac{d'_{1}(w)\nabla u\cdot\nabla w}{u}dx\\ &\quad-\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u})\int_{\Omega}vdx\\ &: = -\int_{\Omega}X_{2}A_{2}X_{2}^{T}dx-\int_{\Omega}Y_{2}B_{2}Y_{2}^{T}dx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u})\int_{\Omega}vdx, \end{aligned} \end{equation} (4.34)

    where X_{2} = (u-\overline{u}, v, w-\overline{w}) and Y_{2} = \left(\frac{\nabla u}{u}, \nabla w\right) , as well as

    \begin{equation} A_{2} = \left( \begin{array}{ccc} \frac{\alpha_{1}}{\gamma_{1}} &\frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right) &0\\ \frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right) &\frac{\alpha_{2}}{\gamma_{2}} &0\\ 0&0&\mu \end{array} \right), B_{2} = \left( \begin{array}{ccc} \frac{\overline{u}d_{1}(w)}{\gamma_{1}} &\frac{\overline{u}d'_{1}(w)}{2\gamma_{1}}\\ \frac{\overline{u}d'_{1}(w)}{2\gamma_{1}}&\frac{D\overline{w}}{w^{2}} \end{array} \right). \end{equation} (4.35)

    It follows from (1.20) that

    \begin{equation} \begin{split} \left|\frac{\alpha_{1}}{\gamma_{1}}\right| > 0 \quad\text{and}\quad\left|\begin{matrix} \frac{\alpha_{1}}{\gamma_{1}} & \frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)\\ \frac{1}{2}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right) & \frac{\alpha_{2}}{\gamma_{2}} \end{matrix}\right| = \frac{\alpha_{1}\alpha_{2}}{\gamma_{1}\gamma_{2}}-\frac{1}{4}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)^{2} > 0 \end{split} \end{equation} (4.36)

    as well as

    \begin{equation} |A_{2}| = \mu\left( \frac{\alpha_{1}\alpha_{2}}{\gamma_{1}\gamma_{2}}-\frac{1}{4}\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)^{2}\right) > 0, \end{equation} (4.37)

    which implies that the matrix A_{2} is positive definite as according to Sylvester's criterion. Similarly, we deduce from (1.23) that

    \begin{equation} \begin{split} \left|\frac{\overline{u}d_{1}(w)}{\gamma_{1}}\right| > 0 \quad\text{and}\quad |B_{2}| = \frac{\overline{u}\overline{w}d_{1}(w)}{\gamma_{1}w^{2}}\left(D-\frac{\overline{u}w^{2}|d'_{1}(w)|^{2}}{4\gamma_{1}\overline{w}d_{1}(w)}\right) > 0, \end{split} \end{equation} (4.38)

    which implies that the matrix B_{2} is positive definite. Thus there exist positive constants \iota_{1} and \iota_{2} such that

    \begin{equation} \begin{split} X_{2}A_{2}X_{2}^{T}\geq \iota_{1}|X_{2}|^{2}\quad\text{and}\quad Y_{2}B_{2}Y_{2}^{T}\geq \iota_{2}|Y_{2}|^{2} \end{split} \end{equation} (4.39)

    for all x\in \Omega and t > 0 . Let \varepsilon_{2}\in (0, \min\{\iota_{1}, \iota_{2}\}) ; we obtain

    \begin{equation} \begin{split} \frac{\text{d}}{\text{dt}}\mathcal{E}_{2}(t)\leq-\varepsilon_{2}\int_{\Omega}|X_{2}|^{2}+|Y_{2}|^{2}dx -\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u})\int_{\Omega}vdx\quad\text{for all}\;\;t > 0, \end{split} \end{equation} (4.40)

    which implies that (4.30) holds. The proof of Lemma 4.7 is complete.

    With the help of Lemma 4.7, we shall give the following stabilization of the semi-trivial steady state (\overline{u}, 0, \overline{w}) for System (1.11).

    Lemma 4.8. Let the assumptions of Theorem 1.3 hold. Then the global bounded solution (u, v, w) of (1.11) converges to the semi-trivial steady state (\overline{u}, 0, \overline{w}) given by (1.24), i.e.,

    \begin{equation} ||u(\cdot,t)-\overline{u}||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}+||w(\cdot,t)-\overline{w}||_{L^{\infty}(\Omega)}\rightarrow 0 \end{equation} (4.41)

    as t\rightarrow \infty .

    Proof. The proof of this lemma is similar to that of Lemma 4.5; here we omit the details.

    Now, we give the convergence rate of the semi-trivial steady state (\overline{u}, 0, \overline{w}) for System (1.11).

    Lemma 4.9. Let the assumptions of Theorem 1.3 hold; then, there exist positive constants C and \lambda such that:

    (a) when \theta = \gamma_{2}\overline{w}-\beta_{2}\overline{u} , then

    \begin{equation} ||u(\cdot,t)-\overline{u}||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}+||w(\cdot,t)-\overline{w}||_{L^{\infty}(\Omega)}\leq C(1+t)^{-\lambda}\;\;\text{for all}\;\; t > T_{2}; \end{equation} (4.42)

    (b) when \theta > \gamma_{2}\overline{w}-\beta_{2}\overline{u} , then

    \begin{equation} ||u(\cdot,t)-\overline{u}||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}+||w(\cdot,t)-\overline{w}||_{L^{\infty}(\Omega)}\leq Ce^{-\lambda t}\;\;\text{for all}\;\; t > T_{2}, \end{equation} (4.43)

    where T_{2} > 0 is some fixed time.

    Proof. Let

    \begin{equation} \mathcal{F}_{2}^{*}(t): = \int_{\Omega}\left(u -\overline{u}\right)^{2}dx+\int_{\Omega}v^{2}dx+ \int_{\Omega}\left(w-\overline{w}\right)^{2}dx, \end{equation} (4.44)

    then it follows from Lemma 4.7 that there exists a \varepsilon_{2} > 0 such that

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{2}(t) \leq-\varepsilon_{2} \mathcal{F}_{2}^{*}(t)-\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u})\int_{\Omega}vdx \;\;\;\text{for all t > 0 }. \end{aligned} \end{equation} (4.45)

    We deduce from Lemma 4.8 that ||u(\cdot, t)-\overline{u}||_{L^{\infty}(\Omega)}+||v(\cdot, t)||_{L^{\infty}(\Omega)}+||w(\cdot, t)-\overline{w}||_{L^{\infty}(\Omega)}\rightarrow 0 as t\rightarrow \infty . Hence, we apply L'Hôpital's rule to obtain

    \begin{equation} \lim\limits_{u\rightarrow \overline{u}}\frac{u-\overline{u}-\overline{u}\ln\frac{u}{\overline{u}}}{(u-\overline{u})^{2}} = \frac{1}{2\overline{u}}, \end{equation} (4.46)

    which implies that there exists a t'_{1} > 0 such that

    \begin{equation} \begin{split} \frac{1}{4\overline{u}}\int_{\Omega}(u-\overline{u})^{2}dx\leq\int_{\Omega}\left(u-\overline{u}-\overline{u}\ln\frac{u}{\overline{u}}\right)dx \leq\frac{3}{4\overline{u}}\int_{\Omega}(u-\overline{u})^{2}dx \end{split} \end{equation} (4.47)

    for all t > t'_{1} . Similarly, we can find t'_{2} > 0 satisfying

    \begin{equation} \begin{split} \frac{1}{4\overline{w}}\int_{\Omega}(w-\overline{w})^{2}dx\leq\int_{\Omega}\left(w-\overline{w}-\overline{w}\ln\frac{w}{\overline{w}}\right)dx \leq\frac{3}{4\overline{w}}\int_{\Omega}(w-\overline{w})^{2}dx \end{split} \end{equation} (4.48)

    for all t > t'_{2} .

    By using the fact that \lim_{s\rightarrow 0}\frac{s}{s^{2}+s} = 1 , it follows from ||v(\cdot, t)||_{L^{\infty}(\Omega)}\rightarrow 0 as t\rightarrow \infty that there exists a t'_{3} > 0 such that

    \begin{equation} \begin{split} \frac{1}{2}\int_{\Omega}v^{2}+vdx\leq\int_{\Omega}vdx\leq\frac{3}{2}\int_{\Omega}v^{2}+vdx \end{split} \end{equation} (4.49)

    for all t > t'_{3} .

    (a) When \theta = \gamma_{2}\overline{w}-\beta_{2}\overline{u} , (4.45) can be turned into

    \begin{equation} \begin{aligned} \frac{d}{d t} \mathcal{E}_{2}(t) \leq-\varepsilon_{2} \mathcal{F}_{2}^{*}(t) \;\;\;\text{for all t > 0 }. \end{aligned} \end{equation} (4.50)

    Let T_{2}: = \max\{t'_{1}, t'_{2}, t'_{3}\} ; by means of the definitions of \mathcal{E}_{2}(t) and \mathcal{F}_{2}^{*}(t) , it follows from the second inequalities in (4.47) and (4.48) that there exist positive constants C_{1} and C_{2} such that

    \begin{equation} \begin{split} \mathcal{E}_{2}(t)&\leq \frac{3}{4\gamma_{1}\overline{u}}\int_{\Omega}(u-\overline{u})^{2}dx+\frac{1}{\gamma_{2}}\int_{\Omega}vdx+\frac{3}{4\overline{w}}\int_{\Omega}(w-\overline{w})^{2}dx\\ &\leq C_{1}\left(\int_{\Omega}(u-\overline{u})^{2}dx\right)^{\frac{1}{2}}+C_{1}\left(\int_{\Omega}v^{2}dx\right)^{\frac{1}{2}}+ C_{1}\left(\int_{\Omega}(w-\overline{w})^{2}dx\right)^{\frac{1}{2}}\\ &\leq C_{2}(\mathcal{F}_{2}^{*}(t))^{\frac{1}{2}}, \end{split} \end{equation} (4.51)

    for all t > T_{2} , where we have used Hölder's inequality and the boundedness of (u, v, w) asserted by Theorem 1.1. Thus, we deduce from (4.50) that

    \begin{equation} \mathcal{E}_{2}'(t)\leq-\frac{\varepsilon_{2}}{C_{2}^{2}}\mathcal{E}_{2}^{2}(t)\quad\text{for all}\;\;t > T_{2}, \end{equation} (4.52)

    which implies

    \begin{equation} \mathcal{E}_{2}(t)\leq \frac{C_{3}}{t-T_{2}}\quad\text{for all}\;\;t > T_{2}, \end{equation} (4.53)

    with some positive constant C_{3} . Hence we infer from the first inequalities in (4.47)–(4.49) that there exists a C_{4} > 0 such that

    \begin{equation} \begin{split} &\int_{\Omega}(u-\overline{u})^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-\overline{w})^{2}dx\\ &\quad\leq C_{4}\mathcal{E}_{2}(t)\leq \frac{C_{3}C_{4}}{t-T_{2}}\;\;\text{for all}\;\;t > T_{2}. \end{split} \end{equation} (4.54)

    It follows from the Gagliardo-Nirenberg inequality in two dimensions, Lemma 4.2 and Lemma 3.2 that there exist positive constants C_{5} and C_{6} such that

    \begin{equation} \begin{split} &||u-\overline{u}||_{L^{\infty}(\Omega)}+||v||_{L^{\infty}(\Omega)}+||w-\overline{w}||_{L^{\infty}(\Omega)}\\ &\leq C_{5}\bigg(||u-\overline{u}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||u-\overline{u}||_{L^{2}(\Omega)}^{\frac{1}{3}} +||v||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||v||_{L^{2}(\Omega)}^{\frac{1}{3}}\\& \quad+||w-\overline{w}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||w-\overline{w}||_{L^{2}(\Omega)}^{\frac{1}{3}}\bigg)\\ &\leq C_{6}\left(\int_{\Omega}(u-\overline{u})^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-\overline{w})^{2}dx\right)^{\frac{1}{6}}\\ &\leq C_{6}(C_{3}C_{4})^{\frac{1}{6}}(t-T_{2})^{-\frac{1}{6}} \end{split} \end{equation} (4.55)

    for all t > T_{2} .

    (b) When \theta > \gamma_{2}\overline{w}-\beta_{2}\overline{u} , let T_{2}: = \max\{t'_{1}, t'_{2}, t'_{3}\} ; by means of the definitions of \mathcal{E}_{2}(t) and \mathcal{F}_{2}^{*}(t) , it follows from the second inequalities in (4.47) and (4.48) that there exists a positive constant C_{7} such that

    \begin{equation} \begin{split} \mathcal{E}_{2}(t)\leq C_{7}\left(\mathcal{F}_{2}^{*}(t)+\int_{\Omega}vdx\right), \end{split} \end{equation} (4.56)

    for all t > T_{2} .

    By combining (4.45) with (4.56), we have

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{2}(t) \leq-\frac{\varepsilon_{2}}{C_{7}} \mathcal{E}_{2}(t)-\frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u}-\gamma_{2}\varepsilon_{2})\int_{\Omega}vdx \;\;\;\text{for all}\; t > T_{2} . \end{aligned} \end{equation} (4.57)

    Since \theta > \gamma_{2}\overline{w}-\beta_{2}\overline{u} , then we can select \varepsilon_{2}\leq \frac{1}{\gamma_{2}}(\theta-\gamma_{2}\overline{w}+\beta_{2}\overline{u}) such that

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{2}(t) \leq-\frac{\varepsilon_{2}}{C_{7}} \mathcal{E}_{2}(t)\;\;\;\text{for all}\; t > T_{2} , \end{aligned} \end{equation} (4.58)

    which means that there exist C_{8} > 0 and C_{9} > 0 satisfying

    \begin{equation} \mathcal{E}_{2}(t)\leq C_{8}e^{-C_{9}(t-T_{2})}\quad\text{for all}\;\;t > T_{2}. \end{equation} (4.59)

    Thus we deduce from the first inequalities in (4.47)–(4.49) that there exists a C_{10} > 0 such that

    \begin{equation} \begin{split} &\int_{\Omega}(u-\overline{u})^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-\overline{w})^{2}dx\\ &\quad\leq C_{10}\mathcal{E}_{2}(t)\leq C_{8}C_{10}e^{-C_{9}(t-T_{2})}\;\;\text{for all}\;\;t > T_{2}. \end{split} \end{equation} (4.60)

    It follows from the Gagliardo-Nirenberg inequality in two dimensions, Lemma 4.2 and Lemma 3.2 that there exist positive constants C_{11} and C_{12} such that

    \begin{equation} \begin{split} &||u-\overline{u}||_{L^{\infty}(\Omega)}+||v||_{L^{\infty}(\Omega)}+||w-\overline{w}||_{L^{\infty}(\Omega)}\\ &\leq C_{11}\bigg(||u-\overline{u}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||u-\overline{u}||_{L^{2}(\Omega)}^{\frac{1}{3}} +||v||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||v||_{L^{2}(\Omega)}^{\frac{1}{3}}\\& \quad+||w-\overline{w}||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||w-\overline{w}||_{L^{2}(\Omega)}^{\frac{1}{3}}\bigg)\\ &\leq C_{12}\left(\int_{\Omega}(u-\overline{u})^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-\overline{w})^{2}dx\right)^{\frac{1}{6}}\\ &\leq C_{12}(C_{8}C_{10})^{\frac{1}{6}}e^{\frac{-C_{9}(t-T_{2})}{6}} \end{split} \end{equation} (4.61)

    for all t > T_{2} . The proof of Lemma 4.9 is complete.

    Proof of Theorem 1.3. The statement of Theorem 1.3 is a direct consequence of Lemma 4.9.

    In this subsection, we are devoted to discussing the asymptotic stability of the prey-only steady state (0, 0, 1) under some suitable parameters conditions. To do this, let us denote the following functionals

    \begin{aligned} \mathcal{E}_{3}(t) = &\frac{1}{\gamma_{1}}\int_{\Omega}udx+\frac{1}{\gamma_{2}}\int_{\Omega}vdx+\int_{\Omega}\left(w-1-\ln w\right)dx\\ \end{aligned}

    and

    \begin{aligned} \mathcal{F}_{3}(t) = \int_{\Omega}u^{2}dx+\int_{\Omega}v^{2}dx+ \int_{\Omega}\left(w-1\right)^{2}dx+\int_\Omega {{{\left| {\nabla w} \right|}^2}}dx, \end{aligned}

    we can derive the following estimates of \mathcal{E}_{3}(t) and \mathcal{F}_{3}(t) .

    Lemma 4.10. Let the conditions of Theorem 1.5 hold. Then there exists a \varepsilon_{3} > 0 independent of t such that

    \begin{equation} \begin{aligned} \mathcal{E}_{3}(t) \geq 0 \end{aligned} \end{equation} (4.62)

    and

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{3}(t) \leq-\varepsilon_{3} \mathcal{F}_{3}(t)-\frac{1}{\gamma_{1}}(\theta-\gamma_{1})\int_{\Omega}udx -\frac{1}{\gamma_{2}}(\theta-\gamma_{2})\int_{\Omega}vdx \;\;\;{for\; all\; t > 0 }. \end{aligned} \end{equation} (4.63)

    Proof. By the similar arguments as in the proofs of Lemma 4.4 and Lemma 4.7, we can derive (4.62) and

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}}\mathcal{E}_{3}(t)& = -\frac{\alpha_{1}}{\gamma_{1}}\int_{\Omega}u^{2}dx -\frac{\alpha_{2}}{\gamma_{2}}\int_{\Omega}v^{2}dx-\mu\int_{\Omega}(w-1)^{2}dx\\&\quad -\left(\frac{\beta_{1}}{\gamma_{1}}+\frac{\beta_{2}}{\gamma_{2}}\right)\int_{\Omega}uvdx-D\int_{\Omega}\frac{|\nabla w|^{2}}{w^{2}}dx\\&\quad-\frac{1}{\gamma_{1}}(\theta-\gamma_{1})\int_{\Omega}udx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2})\int_{\Omega}vdx\\ &\leq-\frac{\alpha_{1}}{\gamma_{1}}\int_{\Omega}u^{2}dx -\frac{\alpha_{2}}{\gamma_{2}}\int_{\Omega}v^{2}dx-\mu\int_{\Omega}(w-1)^{2}dx -\frac{D}{K_{0}^{2}}\int_{\Omega}|\nabla w|^{2}dx\\&\quad-\frac{1}{\gamma_{1}}(\theta-\gamma_{1})\int_{\Omega}udx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2})\int_{\Omega}vdx, \end{aligned} \end{equation} (4.64)

    where we have used the fact that w\leq K_{0} = \max\{||w_{0}||_{L^{\infty}(\Omega)}, 1\} . Let \varepsilon_{3}\in \left(0, \min\{\frac{\alpha_{1}}{\gamma_{1}}, \frac{\alpha_{2}}{\gamma_{2}}, \mu, \frac{D}{K_{0}^{2}}\}\right) ; we obtain

    \begin{equation} \begin{split} \frac{\text{d}}{\text{dt}}\mathcal{E}_{3}(t)\leq-\varepsilon_{3}\mathcal{F}_{3}(t) -\frac{1}{\gamma_{1}}(\theta-\gamma_{1})\int_{\Omega}udx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2})\int_{\Omega}vdx, \end{split} \end{equation} (4.65)

    for all t > 0 . The proof of Lemma 4.10 is complete.

    With the help of Lemma 4.10, we shall give the following stabilization of the prey-only steady state for System (1.11).

    Lemma 4.11. Let the assumptions of Theorem 1.5 hold. Then the global bounded solution of (1.11) converges to the prey-only steady state (0, 0, 1) , i.e.,

    \begin{equation} ||u(\cdot,t)||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}+||w(\cdot,t)-1||_{L^{\infty}(\Omega)}\rightarrow 0 \end{equation} (4.66)

    as t\rightarrow \infty .

    Proof. The proof of this lemma is similar to that of Lemma 4.5; here we omit the details.

    Now, we give the convergence rate of the prey-only steady state (0, 0, 1) for System (1.11).

    Lemma 4.12. Let the assumptions of Theorem 1.5 hold; then there exist positive constants C and \lambda such that:

    (a) when \gamma_{i} = \theta , i = 1, 2 , then

    \begin{equation} ||u(\cdot,t)||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}+||w(\cdot,t)-1||_{L^{\infty}(\Omega)}\leq C(1+t)^{-\lambda}\;\;{for \;all}\;\; t > T_{3}; \end{equation} (4.67)

    (b) when \gamma_{i} < \theta , i = 1, 2 , then

    \begin{equation} ||u(\cdot,t)||_{L^{\infty}(\Omega)}+||v(\cdot,t)||_{L^{\infty}(\Omega)}+||w(\cdot,t)-1||_{L^{\infty}(\Omega)}\leq Ce^{-\lambda t}\;\;{for\; all}\;\; t > T_{3}, \end{equation} (4.68)

    where T_{3} > 0 is some fixed time.

    Proof. Let

    \begin{equation} \mathcal{F}_{3}^{*}(t): = \int_{\Omega}u^{2}dx+\int_{\Omega}v^{2}dx+ \int_{\Omega}\left(w-1\right)^{2}dx, \end{equation} (4.69)

    then it follows from Lemma 4.10 that there exists a \varepsilon_{3} > 0 such that

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{3}(t)\leq-\varepsilon_{3} \mathcal{F}_{3}^{*}(t)-\frac{1}{\gamma_{1}}(\theta-\gamma_{1})\int_{\Omega}udx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2})\int_{\Omega}vdx \;\;\;\text{for all t > 0 }. \end{aligned} \end{equation} (4.70)

    By using the facts that \lim_{s\rightarrow 0}\frac{s}{s^{2}+s} = 1 and \lim_{s\rightarrow 1}\frac{s-1-\ln s}{(s-1)^{2}} = \frac{1}{2} , it follows from ||u(\cdot, t)||_{L^{\infty}(\Omega)}+||v(\cdot, t)||_{L^{\infty}(\Omega)}+||w(\cdot, t)-1||_{L^{\infty}(\Omega)}\rightarrow 0 as t\rightarrow \infty , as asserted in Lemma 4.11 that there exists a T_{3} > 0 such that

    \begin{equation} \begin{split} \frac{1}{2}\int_{\Omega}u^{2}+udx\leq\int_{\Omega}udx\leq\frac{3}{2}\int_{\Omega}u^{2}+udx \end{split} \end{equation} (4.71)

    and

    \begin{equation} \begin{split} \frac{1}{2}\int_{\Omega}v^{2}+vdx\leq\int_{\Omega}vdx\leq\frac{3}{2}\int_{\Omega}v^{2}+vdx \end{split} \end{equation} (4.72)

    as well as

    \begin{equation} \begin{split} \frac{1}{4}\int_{\Omega}(w-1)^{2}dx\leq\int_{\Omega}\left(w-1-\ln w\right)dx \leq\frac{3}{4}\int_{\Omega}(w-1)^{2}dx \end{split} \end{equation} (4.73)

    for all t > T_{3} .

    (a) When \gamma_{i} = \theta , i = 1, 2 , (4.70) can be simplified as

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{3}(t) \leq-\varepsilon_{3} \mathcal{F}_{3}^{*}(t) \;\;\;\text{for all t > 0 }. \end{aligned} \end{equation} (4.74)

    By means of the definitions of \mathcal{E}_{3}(t) and \mathcal{F}_{3}^{*}(t) , it follows from the second inequality in (4.73) that there exist positive constants C_{1} and C_{2} such that

    \begin{equation} \begin{split} \mathcal{E}_{3}(t)&\leq \frac{1}{\gamma_{1}}\int_{\Omega}udx+\frac{1}{\gamma_{2}}\int_{\Omega}vdx+\frac{3}{4}\int_{\Omega}(w-1)^{2}dx\\ &\leq C_{1}\left(\int_{\Omega}u^{2}dx\right)^{\frac{1}{2}}+C_{1}\left(\int_{\Omega}v^{2}dx\right)^{\frac{1}{2}}+ C_{1}\left(\int_{\Omega}(w-1)^{2}dx\right)^{\frac{1}{2}}\\ &\leq C_{2}(\mathcal{F}_{3}^{*}(t))^{\frac{1}{2}}, \end{split} \end{equation} (4.75)

    for all t > T_{3} , where we have used Hölder's inequality and the boundedness of (u, v, w) asserted by Theorem 1.1. Thus we deduce from (4.74) that

    \begin{equation} \mathcal{E}_{3}'(t)\leq-\frac{\varepsilon_{3}}{C_{2}^{2}}\mathcal{E}_{3}^{2}(t)\quad\text{for all}\;\;t > T_{3}, \end{equation} (4.76)

    which implies

    \begin{equation} \mathcal{E}_{3}(t)\leq \frac{C_{3}}{t-T_{3}}\quad\text{for all}\;\;t > T_{3}, \end{equation} (4.77)

    with some positive constant C_{3} . Hence we infer from the first inequalities in (4.71)–(4.73) that there exists a C_{4} > 0 such that

    \begin{equation} \begin{split} &\int_{\Omega}u^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-1)^{2}dx\\ &\quad\leq C_{4}\mathcal{E}_{3}(t)\leq \frac{C_{3}C_{4}}{t-T_{3}}\;\;\text{for all}\;\;t > T_{3}. \end{split} \end{equation} (4.78)

    It follows from the Gagliardo-Nirenberg inequality in two dimensions, Lemma 4.2 and Lemma 3.2 that there exist positive constants C_{5} and C_{6} such that

    \begin{equation} \begin{split} &||u||_{L^{\infty}(\Omega)}+||v||_{L^{\infty}(\Omega)}+||w-1||_{L^{\infty}(\Omega)}\\ &\leq C_{5}\bigg(||u||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||u||_{L^{2}(\Omega)}^{\frac{1}{3}} +||v||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||v||_{L^{2}(\Omega)}^{\frac{1}{3}}\\& \quad+||w-1||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||w-1||_{L^{2}(\Omega)}^{\frac{1}{3}}\bigg)\\ &\leq C_{6}\left(\int_{\Omega}u^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-1)^{2}dx\right)^{\frac{1}{6}}\\ &\leq C_{6}(C_{3}C_{4})^{\frac{1}{6}}(t-T_{3})^{-\frac{1}{6}} \end{split} \end{equation} (4.79)

    for all t > T_{3} .

    (b) When \gamma_{i} < \theta , i = 1, 2 , by means of the definitions of \mathcal{E}_{3}(t) and \mathcal{F}_{3}^{*}(t) , it follows from the second inequalities in (4.71)–(4.73) that there exists a positive constant C_{7} such that

    \begin{equation} \begin{split} \mathcal{E}_{3}(t)\leq C_{7}\left(\mathcal{F}_{3}^{*}(t)+\int_{\Omega}udx+\int_{\Omega}vdx\right), \end{split} \end{equation} (4.80)

    for all t > T_{3} .

    By combining (4.70) with (4.80), we derive

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{3}(t)\leq-\frac{\varepsilon_{3}}{C_{7}} \mathcal{E}_{3}(t)-\frac{1}{\gamma_{1}}(\theta-\gamma_{1}-\varepsilon_{3}\gamma_{1})\int_{\Omega}udx-\frac{1}{\gamma_{2}}(\theta-\gamma_{2} -\varepsilon_{3}\gamma_{2})\int_{\Omega}vdx \end{aligned} \end{equation} (4.81)

    for all t > 0 . Since \gamma_{i} < \theta , i = 1, 2 , we can select \varepsilon_{3}\leq \min\left\{\frac{1}{\gamma_{1}}(\theta-\gamma_{1}), \frac{1}{\gamma_{2}}(\theta-\gamma_{2})\right\} such that

    \begin{equation} \begin{aligned} \frac{\text{d}}{\text{dt}} \mathcal{E}_{3}(t) \leq-\frac{\varepsilon_{3}}{C_{7}} \mathcal{E}_{3}(t)\;\;\;\text{for all t > T_{3} }, \end{aligned} \end{equation} (4.82)

    which means that there exist C_{8} > 0 and C_{9} > 0 satisfying

    \begin{equation} \mathcal{E}_{3}(t)\leq C_{8}e^{-C_{9}(t-T_{3})}\quad\text{for all}\;\;t > T_{3}. \end{equation} (4.83)

    Thus we deduce from the first inequalities in (4.71)–(4.73) that there exists a C_{10} > 0 such that

    \begin{equation} \begin{split} &\int_{\Omega}u^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-1)^{2}dx\\ &\quad\leq C_{10}\mathcal{E}_{2}(t)\leq C_{8}C_{10}e^{-C_{9}(t-T_{3})}\;\;\text{for all}\;\;t > T_{3}. \end{split} \end{equation} (4.84)

    It follows from the Gagliardo-Nirenberg inequality in two dimensions, Lemma 4.2 and Lemma 3.2 that there exist positive constants C_{11} and C_{12} such that

    \begin{equation} \begin{split} &||u||_{L^{\infty}(\Omega)}+||v||_{L^{\infty}(\Omega)}+||w-1||_{L^{\infty}(\Omega)}\\ &\leq C_{11}\bigg(||u||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||u||_{L^{2}(\Omega)}^{\frac{1}{3}} +||v||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||v||_{L^{2}(\Omega)}^{\frac{1}{3}}\\& \quad+||w-1||_{W^{1,4}(\Omega)}^{\frac{2}{3}}||w-1||_{L^{2}(\Omega)}^{\frac{1}{3}}\bigg)\\ &\leq C_{12}\left(\int_{\Omega}u^{2}dx+\int_{\Omega}v^{2}dx+\int_{\Omega}(w-1)^{2}dx\right)^{\frac{1}{6}}\\ &\leq C_{12}(C_{8}C_{10})^{\frac{1}{6}}e^{\frac{-C_{9}(t-T_{3})}{6}} \end{split} \end{equation} (4.85)

    for all t > T_{3} . The proof of Lemma 4.12 is complete.

    Proof of Theorem 1.5. The statement of Theorem 1.5 is a direct consequence of Lemma 4.12.

    The author is grateful to the editor and three anonymous referees for their valuable comments, which greatly improved the exposition of our paper. The author is also deeply grateful to Professor Renjun Duan for the support and friendly hospitality in CUHK. The work was partially supported by the National Natural Science Foundation of China (Grant Nos: 11601053, 11526042), Natural Science Foundation of Chongqing (Grant No. cstc2019jcyj-msxmX0082), Chongqing Municipal Education Commission Science and Technology Research Project (Grant No. KJZD-K202200602), China-South Africa Young Scientist Exchange Project in 2020, The Hong Kong Scholars Program (Grant Nos: XJ2021042, 2021-005) and Young Hundred Talents Program of CQUPT from 2022–2024.

    The author declares that there is no conflict of interest.



    [1] J. A. DiMasi, R. W. Hansen, H. G. Grabowski, The price of innovation: new estimates of drug development costs, J Health Econ., 22 (2003), 151–185. https://doi.org/10.1016/s0167-6296(02)00126-1 doi: 10.1016/s0167-6296(02)00126-1
    [2] M. A. Yildirim, K. Goh, M. E. Cusick, A. Barabas, M. Vidal, Drug-target network, Nat. Biotechnol., 25 (2007), 1119–1126. https://doi.org/10.1038/nbt1338 doi: 10.1038/nbt1338
    [3] T. T. Ashburn, K. B. Thor, Drug repositioning: identifying and developing new uses for existing drugs, Nat. Rev. Drug Discov., 3 (2004), 673–683. https://doi.org/10.1038/nrd1468 doi: 10.1038/nrd1468
    [4] F. Cheng, C. Liu, J. Jiang, W. Lu, W. Li, G. Liu, et al., Prediction of drug-target interactions and drug repositioning via network-based inference, PLoS Comput. Biol., 8 (2012), 1002503. https://doi.org/10.1371/journal.pcbi.1002503 doi: 10.1371/journal.pcbi.1002503
    [5] Y. Yamanishi, M. Kotera, M. Kanehisa, S. Goto, Drug-target interaction prediction from chemical, genomic and pharmacological data in an integrated framework, Bioinformatics, 26 (2010), 246–254. https://doi.org/10.1093/bioinformatics/btq176 doi: 10.1093/bioinformatics/btq176
    [6] A. C. Cheng, R. G. Coleman, K. T. Smyth, Q. Cao, P. Soulard, D. R. Caffrey, et al., Structure-based maximal affinity model predicts small-molecule druggability, Nat. Biotechnol., 25 (2007), 71–75. https://doi.org/10.1038/nbt1273 doi: 10.1038/nbt1273
    [7] H. Alonso, A. A. Bliznyuk, J. E. Gready, Combining docking and molecular dynamic simulations in drug design, Med. Res. Rev., 26 (2006), 531–568. https://doi.org/10.1002/med.20067 doi: 10.1002/med.20067
    [8] K. Wang, J. Sun, S. Zhou, C. Wan, S. Qin, C. Li, et al., Prediction of drug–target interactions for drug repositioning only based on genomic expression similarity, PLoS Comput. Biol., 9 (2013), e1003315. https://doi.org/10.1371/journal.pcbi.1003315 doi: 10.1371/journal.pcbi.1003315
    [9] A.Tropsha, Best practices for QSAR model development, validation, and exploitation, Mol. Inform., 29(2010), 476–488. http://doi.org/10.1002/minf.201000061 doi: 10.1002/minf.201000061
    [10] D. Rognan, Chemogenomic approaches to rational drug design, Br. J. Pharmacol., 152 (2007), 38–52. https://doi.org/10.1038/sj.bjp.0707307 doi: 10.1038/sj.bjp.0707307
    [11] A. C. A. Nascimento, R. B. C. Prudêncio, I. G. Cost, A multiple kernel learning algorithm for drug-target interaction prediction, BMC Bioinformatics, 17 (2016), 1–16, https://doi.org/10.1186/s12859-016-0890-3 doi: 10.1186/s12859-016-0890-3
    [12] T. He, M. Heidemeyer, F. Ban, A. Cherkasov, M. Ester, SimBoost: a read-across approach for predicting drug–target binding affinities using gradient boosting machines, J. Cheminform., 9 (2017), 1–14. https://doi.org/10.1186/s13321-017-0209-z doi: 10.1186/s13321-017-0209-z
    [13] K. Y. Gao, A. Fokoue, H. Luo, A. Iyengar, S. Dey, P. Zhang, Interpretable drug target prediction using deep neural representation, in IJCAI International Joint Conferences on Artificial Intelligence Organization, (2018), 3371–3377. https://doi.org/10.24963/ijcai.2018/468
    [14] M. Tsubaki, K. Tomii, J. Sese, Compound–protein interaction prediction with end-to-end learning of neural networks for graphs and sequences, Bioinformatics 35 (2019), 309–318. https://doi.org/10.1093/bioinformatics/bty535 doi: 10.1093/bioinformatics/bty535
    [15] Y. Luo, X. Zhao, J. Zhou, J. Yang, Y. Zhang, W. Kuang, et al., A network integration approach for drug-target interaction prediction and computational drug repositioning from heterogeneous information, Nat. Commun., 8 (2017), 573. https://doi.org/10.1038/s41467-017-00680-8 doi: 10.1038/s41467-017-00680-8
    [16] R. S. Olayan, H. Ashoor, V. B. Bajic, DDR: efficient computational method to predict drug–target interactions using graph mining and machine learning approaches, Bioinformatics, 34 (2018), 1164–1173. https://doi.org/10.1093/bioinformatics/btx731 doi: 10.1093/bioinformatics/btx731
    [17] W. Zhang, W. Lin, D. Zhang, S. Wang, J. Shi, Y. Niu, Recent advances in the machine learning-based drug-target interaction prediction, Curr. Drug Metab., 20 (2019), 194–202. https://doi.org/10.2174/1389200219666180821094047 doi: 10.2174/1389200219666180821094047
    [18] M. Thafar, A. B. Raies, S. Albaradei, M. Essack, V. B. Bajic, Comparison study of computational prediction tools for drug–target binding affinities, Front. Chem., 7 (2019), 782. https://doi.org/10.3389/fchem.2019.00782 doi: 10.3389/fchem.2019.00782
    [19] S. Anusuya, M. Kesherwani, K. V. Priya, A. Vimala, G. Shanmugam, D. Velmurugan, et al., Drug–target interactions: prediction methods and applications, Curr. Protein Pept. Sci., 19 (2018), 537–561. https://doi.org/10.2174/1389203718666161108091609 doi: 10.2174/1389203718666161108091609
    [20] K. Sachdev, M. K. Gupta, A comprehensive review of feature based methods for drug target interaction prediction, J Biomed Inform., 93 (2019), 103159. https://doi.org/10.1016/j.jbi.2019.103159 doi: 10.1016/j.jbi.2019.103159
    [21] W. Chen, L. Kurgan, Survey of similarity-based prediction of drug-protein interactions, Curr. Med. Chem., 27 (2020), 5856–5886. https://doi.org/10.2174/0929867326666190808154841 doi: 10.2174/0929867326666190808154841
    [22] K. Abbasi, P. Razzaghi, A. Poso, S. Ghanbari-Ara, A. Masoudi-Nejad, Deep learning in drug target interaction prediction, Curr. Med. Chem., 28 (2021), 2100–2113. https://doi.org/10.2174/0929867327666200907141016 doi: 10.2174/0929867327666200907141016
    [23] Q. Bai, J. Ma, S. Liu, T. Xu, A. J. Banegas-Luna, H. Pérez-Sánchez, et al., WADDAICA: a webserver for aiding protein drug design by artificial intelligence and classical algorithm, Comput. Struct. Biotechnol. J., 19 (2021), 3573–3579. https://doi.org/10.1016/j.csbj.2021.06.017 doi: 10.1016/j.csbj.2021.06.017
    [24] Q. Bai, S. Tan, T. Xu, H. Liu, J. Huang, X. Yao, MolAICal: a soft tool for 3D drug design of protein targets by artificial intelligence and classical algorithm, Brief. Bioinform., 22 (2021), 161. https://doi.org/10.1093/bib/bbaa161 doi: 10.1093/bib/bbaa161
    [25] Q. Bai, S. Liu, Y. Tian, T. Xu, A. J. Banegas‐Luna, H. Pérez‐Sánchez, et al., Application advances of deep learning methods for de novo drug design and molecular dynamics simulation, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 12 (2022), 1581. https://doi.org/10.1002/wcms.1581 doi: 10.1002/wcms.1581
    [26] M. Koutrouli, E. Karatzas, D. Paez-Espino, G. A. Pavlopoulos, A guide to conquer the biological network era using graph theory, Front. Bioeng. Biotechnol., 8 (2020), 34. http://doi.org/10.3389/fbioe.2020.00034 doi: 10.3389/fbioe.2020.00034
    [27] Y. Tabei, M. Kotera, R. Sawada, Y. Yamanishi, Network-based characterization of drug-protein interaction signatures with a space-efficient approach, BMC Syst. Biol., 39 (2019). https://doi.org/10.1186/s12918-019-0691-1 doi: 10.1186/s12918-019-0691-1
    [28] W. Wang, H. Lv, Y. Zhao, D. Liu, Y. Wang, Y. Zhang, DLS: A link prediction method based on network local structure for predicting drug-protein interactions, Front. Bioeng. Biotechnol., 8 (2020). https://doi.org/10.3389/fbioe.2020.00330 doi: 10.3389/fbioe.2020.00330
    [29] Z. Wu, W. Li, G. Liu, Y. Tang, Network-based methods for prediction of drug-target interactions, Front. Pharmacol., 9 (2018), 1134. https://doi.org/10.3389/fphar.2018.01134 doi: 10.3389/fphar.2018.01134
    [30] K. Abbas, A. Abbasi, S. Dong, L. Niu, L. Yu, B. Chen, et al., Application of network link prediction in drug discovery, BMC Bioinformatics, 187 (2021). https://doi.org/10.1186/s12859-021-04082-y doi: 10.1186/s12859-021-04082-y
    [31] M. A. Thafar, R. S. Olayan, H. Ashoor, S. Albaradei, V. B. Bajic, X. Gao, et al., DTiGEMS+: drug–target interaction prediction using graph embedding, graph mining, and similarity-based techniques, J. Cheminform., 12 (2020), 1–17. https://doi.org/10.1186/s13321-020-00447-2 doi: 10.1186/s13321-020-00447-2
    [32] L. Wang, Z. You, X. Chen, S. Xia, F. Liu, X. Yan, et al., Computational methods for the prediction of drug– target interactions from drug fingerprints and protein sequences by stacked auto-encoder deep neural network, in International Symposium on Bioinformatics Research and Applications, 46–58. https://doi.org/10.1007/978-3-319-59575-7_5
    [33] B. Agyemang, W. P. Wu, M. Y. Kpiebaareh, Z. Lei, E. Nanor, L. Chen, Multi-view self-attention for interpretable drug–target interaction prediction, J. Biomed. Inform., 110 (2020), 103547. https://doi.org/10.1016/j.jbi.2020.103547 doi: 10.1016/j.jbi.2020.103547
    [34] P. Zhang, Z. Wei, C. Che, B. Jin, DeepMGT-DTI: Transformer network incorporating multilayer graph information for Drug-Target interaction prediction, Comput. Biol. Med., 142 (2022), 105214. https://doi.org/10.1016/j.compbiomed.2022.105214 doi: 10.1016/j.compbiomed.2022.105214
    [35] C. Chen, H. Shi, Z Jiang, A. Salhi, R. Chen, X. Cui, et al., DNN-DTIs: Improved drug-target interactions prediction using XGBoost feature selection and deep neural network, Comput. Biol. Med., 136 (2021), 104676. https://doi.org/10.1016/j.compbiomed.2021.104676 doi: 10.1016/j.compbiomed.2021.104676
    [36] X. Zeng, S. Zhu, W. Lu, Z. Liu, J. Huang, Y Zhou, et al., Target identification among known drugs by deep learning from heterogeneous networks, Chem. Sci., 11 (2020), 1775–1797. https://doi.org/10.1039/C9SC04336E doi: 10.1039/C9SC04336E
    [37] B. Wang, A. M. Mezlini, F. Demir, M. Fiume, Z. Tu, M. Brudno, et al., Similarity network fusion for aggregating data types on a genomic scale, Nat. Methods, 11 (2014), 333–337. https://doi.org/10.1038/nmeth.2810 doi: 10.1038/nmeth.2810
    [38] F. Xia, J. Liu, H. Nie, Y. Fu, L. Wan, X. Kong, Random walks: A review of algorithms and applications, IEEE Trans. Emerging Top. Comput. Intell., 4 (2019), 95–107. https://doi.org/10.1109/TETCI.2019.2952908 doi: 10.1109/TETCI.2019.2952908
    [39] A. Ghulam, X. Lei, M. Guo, C. Bian, Disease-pathway association prediction based on random walks with restart and PageRank, IEEE Access, 8 (2020), 72021–72038. https://doi.org/10.1109/ACCESS.2020.2987071 doi: 10.1109/ACCESS.2020.2987071
    [40] S. Li, L. Jiang, J. Tang, N. Gao, F. Guo, Kernel fusion method for detecting cancer subtypes via selecting relevant expression data, Front. Genet., 11 (2020), 979. https://doi.org/10.3389/fgene.2020.00979 doi: 10.3389/fgene.2020.00979
    [41] R. Qin, L. Duan, H. Zheng, J. Li-Ling, K. Song, Y. Zhang, An ontology-independent representation learning for similar disease detection based on multi-layer similarity network, IEEE/ACM Trans. Comput. Biol. Bioinform., 18 (2021), 183–193. https://doi.org/10.1109/tcbb.2019.2941475 doi: 10.1109/tcbb.2019.2941475
    [42] H. I. Lee, H. Nam, Identification of drug-target interaction by a random walk with restart method on an interactome network, BMC Bioinformatics, 208 (2018), 9–18. https://doi.org/10.1186/s12859-018-2199-x doi: 10.1186/s12859-018-2199-x
    [43] T. N. Jarada, J. G. Rokne, R. Alhajj, SNF-NN: computational method to predict drug-disease interactions using similarity network fusion and neural networks, BMC Bioinformatics, 28 (2021), 1–20. https://doi.org/10.1186/s12859-020-03950-3 doi: 10.1186/s12859-020-03950-3
    [44] D. V. Anand, Q. Xu, J. Wee, K. Xia, T. C. Sum, Topological feature engineering for machine learning based halide perovskite materials design, Npj Comput. Mater., 203 (2022), 1–8. https://doi.org/10.1038/s41524-022-00883-8 doi: 10.1038/s41524-022-00883-8
    [45] M. Wang, Z. Cang, G. W. Wei, A Topology-based network tree for the prediction of protein-protein binding affinity changes following mutation, Nat. Mach. Intell., 2 (2020), 116–123. https://doi.org/10.1038/s42256-020-0149-6 doi: 10.1038/s42256-020-0149-6
    [46] S. Fan, J. Zhu, X. Han, C. Shi, L. Hu, B. Ma, et al., Metapath-guided heterogeneous graph neural network for intent recommendation, in Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, (2019), 2478–2486. https://doi.org/10.1145/3292500.3330673
    [47] X. Fu, J. Zhang, Z. Meng, I. King, MAGNN: Metapath aggregated graph neural network for heterogeneous graph embedding, in Proceedings of The Web Conference, (2020), 2331–234. https://doi.org/10.1145/3366423.3380297
    [48] Z. Lu, Y. Wang, M. Zeng, M. Li, HNEDTI: Prediction of drug-target interaction based on heterogeneous network embedding, in 2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2019,211–214. https://doi.org/10.1109/BIBM47256.2019.8983181
    [49] M. Samizadeh, B. Minaei-Bidgoli, Drug-target interaction prediction by Metapath2vec node embedding in heterogeneous network of interactions, Int. J. Artif. Intell. Tools, 29 (2020), 2050001. https://doi.org/10.1142/S0218213020500013 doi: 10.1142/S0218213020500013
    [50] X. Liu, H. Feng, J. Wu, K. Xia, Persistent spectral hypergraph-based machine learning (PSH-ML) for protein-ligand binding affinity prediction, Brief. Bioinform., 22 (2021), 127. https://doi.org/10.1093/bib/bbab127 doi: 10.1093/bib/bbab127
    [51] X. Liu, X. Wang, J. Wu, K. Xia, Hypergraph-based persistent cohomology (HPC) for molecular representations in drug design, Brief. Bioinform., 22 (2021), 411. https://doi.org/10.1093/bib/bbaa411 doi: 10.1093/bib/bbaa411
    [52] G. A. Petsko, Guilt by association, Genome Biol., 104 (2009), 1–3. https://doi.org/10.1186/gb-2009-10-4-104 doi: 10.1186/gb-2009-10-4-104
    [53] H. Wang, H. Zheng, J. Wang, C. Wang, F. Wu, Integrating omics data with a multiplex network-based approach for the identification of cancer subtypes, IEEE Trans. Nanobioscience, 15 (2016), 335–342. https://doi.org/10.1109/tnb.2016.2556640 doi: 10.1109/tnb.2016.2556640
    [54] D. S. Wishart, Y. D. Feunang, A. C. Guo, E. J. Lo, A. Marcu, J. R. Grant, et al., DrugBank 5.0: a major update to the DrugBank database for 2018, Nucleic Acids Res., 46 (2018), 1074–1082. https://doi.org/10.1093/nar/gkx1037 doi: 10.1093/nar/gkx1037
    [55] D. Weininger, SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules, J. Chem. Inf. Comput. Sci, 28 (1988), 31–36. https://doi.org/10.1021/ci00057a005 doi: 10.1021/ci00057a005
    [56] C. Steinbeck, Y. Han, S. Kuhn, O. Horlacher, E. Luttmann, E. Willighagen, Recent developments of the chemistry development kit (CDK)-an opensource java library for chemo-and bioinformatics, Curr. Pharm. Des., 12 (2006), 2111–2120. https://doi.org/10.1021/ci025584y doi: 10.1021/ci025584y
    [57] A. Rácz, D. Bajusz, K. Héberger, Life beyond the Tanimoto coefficient: similarity measures for interaction fingerprints, J. Cheminform., 10 (2018), 1–12. https://doi.org/10.1186/s13321-018-0302-y doi: 10.1186/s13321-018-0302-y
    [58] M. Kuhn, M. Campillos, I. Letunic, L. J. Jensen, P. Bork, A side effect resource to capture phenotypic effects of drugs, Mol. Syst. Biol., 6 (2010), 343. https://doi.org/10.1038/msb.2009.98 doi: 10.1038/msb.2009.98
    [59] F. Zhu, TTD, 2009. Available from: https://bio.tools/ttd
    [60] UniProt consortium, Uniprot, 2023.Available from: http://www.uniprot.org/
    [61] T. F. Smith, M. S. Waterman, Identification of common molecular subsequences, J Mol. Biol., 147 (1981), 195–197. https://doi.org/10.1016/0022-2836(81)90087-5 doi: 10.1016/0022-2836(81)90087-5
    [62] A. P. Davis, C. J. Grondin, R. J. Johnson, D. Sciaky, B. L. King, R. McMorran, et al., The Comparative toxicogenomics database: update 2017, Nucleic Acids Res., 45 (2017), 972–978. https://doi.org/10.1093/nar/gkw838 doi: 10.1093/nar/gkw838
    [63] Gene Ontology, The Gene Ontology Resource, 2023. Available from: http://www.geneontology.org/
    [64] J. A. Blake, M. Dolan, H. Drabkin, D. P. Hill, L. Ni, D.Sitnikov, et al., Gene Ontology annotations and resources, Nucleic Acids Res., 41 (2013), 530–535. https://doi.org/10.1093/nar/gks1050 doi: 10.1093/nar/gks1050
    [65] G. Yu, F. Li, Y. Qin, X. Bo, Y. Wu, S. Wang, GOSemSim: an R package for measuring semantic similarity among GO terms and gene products, Bioinformatics, 26 (2010), 976–978. https://doi.org/10.1093/bioinformatics/btq064 doi: 10.1093/bioinformatics/btq064
    [66] S. S. Deepika, T. V. Geetha, A meta-learning framework using representation learning to predict drug-drug interaction, J. Biomed. Inform., 84 (2018), 136–147. https://doi.org/10.1016/j.jbi.2018.06.015 doi: 10.1016/j.jbi.2018.06.015
    [67] L. Huang, H. Luo, S. Li, F. Wu, J. Wang, Drug-drug similarity measure and its applications, Brief. Bioinform., 22 (2021), 265. https://doi.org/10.1093/bib/bbaa265 doi: 10.1093/bib/bbaa265
    [68] T. van Laarhoven, S. B. Nabuurs, E. Marchiori, Gaussian interaction profile kernels for predicting drug-target interaction, Bioinformatics, 27 (2011), 3036–3043. https://doi.org/10.1093/bioinformatics/btr500 doi: 10.1093/bioinformatics/btr500
    [69] H. Ye, K. Tang, L. Yang, Z. Cao, Y. Li, Study of drug function based on similarity of pathway fingerprint, Protein Cell, 3 (2012), 132–139. https://doi.org/10.1007/s13238-012-2011-z doi: 10.1007/s13238-012-2011-z
    [70] M. Kanehisa, M. Furumichi, M. Tanabe, Y. Sato, K. Morishima, KEGG: new perspectives on genomes, pathways, diseases and drugs, Nucleic Acids Res., 45 (2017), 353–361. https://doi.org/10.1093/nar/gkw1092 doi: 10.1093/nar/gkw1092
    [71] L. R. Sancho, BioCor: Functional similarities. Available from: https://llrs.github.io/BioCor/
    [72] A. Herbert, MaxCluster, 2023. Available from: http://www.sbg.bio.ic.ac.uk/~maxcluster/
    [73] National Science Foundation, US Department of Energy, National Cancer Institute, National Institute of Allergy and Infectious Diseases, National Institute of General Medical Sciences of the National Institutes of Health, RCSB Protein Data Bank (RCSB PDB), 2023. Available from: https://www.rcsb.org/
    [74] I. Kufareva, R. Abagyan, Methods of protein structure comparison, Homol. Model., 857 (2012), 231–257. https://doi.org/10.1007/978-1-61779-588-6_10 doi: 10.1007/978-1-61779-588-6_10
    [75] X. Wu, E. Pang, K. Lin, Z. M. Pei. Improving the measurement of semantic similarity between gene ontology terms and gene products: Insights from an edge- and IC-Based hybrid method, PLoS ONE, 8 (2013), 66745. https://doi.org/10.1371/journal.pone.0066745 doi: 10.1371/journal.pone.0066745
    [76] M. Shatnawi, Review of recent protein-protein interaction techniques, in Emerging Trends in Computational Biology, Bioinformatics and Systems Biology, (2015), 99–121. http://doi.org/10.1016/B978-0-12-802508-6.00006-5
    [77] C. Menichelli, O. Gascuel, L. Bréhélin, Improving pairwise comparison of protein sequences with domain co-occurrence, PLoS Comput. Biol., 14 (2018), 1005889. https://doi.org/10.1371/journal.pcbi.1005889 doi: 10.1371/journal.pcbi.1005889
    [78] R. D. Finn, A. Bateman, J. Clements, P. Coggil, R. Y. Eberhardt, S. R. Eddy, et al., Pfam: the protein families database, Nucleic Acids Res, 42 (2014), 222–230. https://doi.org/10.1093/nar/gkt1223 doi: 10.1093/nar/gkt1223
    [79] Y. M. Omar, P. Plapper, A survey of information entropy metrics for complex networks, Entropy, 22 (2020), 1417. https://doi.org/10.3390/e22121417 doi: 10.3390/e22121417
    [80] A. P. Riascos, J. L. Mateos, Random walks on weighted networks: Exploring local and non-local navigation strategies, 2019, preprint, arXiv: 1901.05609.
    [81] Chris McCormick, Word2vec tutorial-the skip-gram model, 2016. Available from: http://mccormickml.com/2016/04/19/word2vec-tutorial-the-skip-gram-model/
    [82] B. Wang, A. Wang, F. Chen, Y. Wang, C. C. J. Kuo, Evaluating word embedding models: methods and experimental results, APSIPA Trans. Signal Inf. Process., 8 (2019). https://doi.org/10.1017/ATSIP.2019.12 doi: 10.1017/ATSIP.2019.12
    [83] Y. Li, M. Yang, Z. Zhang, A survey of multi-view representation learning, IEEE Trans. Knowl. Data Eng., 31 (2018), 1863–1883. https://doi.org/10.1109/TKDE.2018.2872063 doi: 10.1109/TKDE.2018.2872063
    [84] K. Sun, L. Wang, B. Xu, W. Zhao, S. W. Teng, F. Xia, Network representation learning: From traditional feature learning to deep learning, IEEE Access, 8 (2020), 205600–205617. https://doi.org/10.1016/j.jfds.2017.05.001 doi: 10.1016/j.jfds.2017.05.001
    [85] A. Grover, J. Leskovec, node2vec: Scalable feature learning for networks, in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, (2016), 855–864. https://doi.org/10.1145/2939672.2939754
    [86] Y. Dong, N. V. Chawla, A. Swami, metapath2vec: Scalable representation learning for heterogeneous networks, in Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, (2017), 135–144. https://doi.org/10.1145/3097983.3098036
    [87] B. Perozzi, R. Al-Rfou, S. Skiena, Deepwalk: Online learning of social representations, in Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining, (2014), 701–710. https://doi.org/10.1145/2623330.2623732
    [88] P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Liò, Y. Bengio, Graph attention networks, in ICLR, (2018).
    [89] G. Douzas, F. Bacao, F. Last, Improving imbalanced learning through a heuristic oversampling method based on k-means and SMOTE, Inf. Sci., 465 (2018), 1–20. https://doi.org/10.1016/j.ins.2018.06.056 doi: 10.1016/j.ins.2018.06.056
    [90] A. Creswell, T. White, V. Dumoulin, K. Arulkumaran, B. Sengupta. A. A. Bharath, Generative adversarial networks: An overview, IEEE Signal Proc. Mag., 35 (2018), 53–65. https://doi.org/10.1109/MSP.2017.2765202 doi: 10.1109/MSP.2017.2765202
    [91] J. Peng, J. Li, X. Shang, A learning-based method for drug-target interaction prediction based on feature representation learning and deep neural network, BMC Bioinformatics, 21 (2020), 1–13. https://doi.org/10.1186/s12859-020-03677-1 doi: 10.1186/s12859-020-03677-1
    [92] Z. Zheng, N. Q. K. Le, M. C. H. Chua, MaskDNA-PGD: An innovative deep learning model for detecting DNA methylation by integrating mask sequences and adversarial PGD training as a data augmentation method, Chemom. Intell. Lab. Syst., 232 (2023), 104715. https://doi.org/10.1016/j.chemolab.2022.104715 doi: 10.1016/j.chemolab.2022.104715
    [93] J. N. Sua, S. Y. Lim, M. H. Yulius, X. Su, E. K. Y. Yapp, N. Q. K. Le, et al., Incorporating convolutional neural networks and sequence graph transform for identifying multilabel protein Lysine PTM sites, Chemom. Intell. Lab. Syst., 206 (2020), 104171. https://doi.org/10.1016/j.chemolab.2020.104171 doi: 10.1016/j.chemolab.2020.104171
    [94] Y. Yamanishi, M. Araki, A. Gutteridge, W. Honda, M. Kanehisa, Prediction of drug-target interaction networks from the integration of chemical and genomic spaces, Bioinformatics, 24 (2008), 232–240. https://doi.org/10.1093/bioinformatics/btn162 doi: 10.1093/bioinformatics/btn162
    [95] Q. Kuang, X. Xu, R. Li, Y. Dong, Y. Li, Z. Huang, et al., An eigenvalue transformation technique for predicting drug-target interaction, Sci. Rep., 5 (2015), 13867. https://doi.org/10.1038/srep13867 doi: 10.1038/srep13867
    [96] J. Davis, M. Goadrich, The relationship between precision-recall and ROC Curves, in Proceedings of the 23rd international conference on machine learning, (2006), 233–240. https://doi.org/10.1145/1143844.1143874
    [97] T. Nguyen, T. H. Nguyen, T. N. H. Pham, N. T. Huy, M. V. Bay, M. Q. Pham, et al., Autodockvina adopts more accurate binding poses but autodock4 forms better binding affinity, J. Chem. Inf. Model., 60 (2019), 204–211. https://doi.org/10.1021/acs.jcim.9b00778 doi: 10.1021/acs.jcim.9b00778
    [98] PunMed, National Library of MedicineNational Center for Biotechnology Information, 2023. Available from: https://pubmed.ncbi.nlm.nih.gov/
    [99] European Molecular Biology Laboratory, ChEMBL, 2023. Available from: https://www.ebi.ac.uk/chembl/
    [100] N. Rohani, C. Eslahchi, Drug-drug interaction predicting by neural network using integrated similarity, Sci. Rep., 9 (2019), 1–11. https://doi.org/10.1038/s41598-019-50121-3 doi: 10.1038/s41598-019-50121-3
    [101] R. Zhou, Z. Lu, H. Luo, J. Xiang, M. Zeng, M. Li, NEDD: a network embedding based method for predicting drug-disease associations, BMC Bioinformatics, 387 (2020), 1–12. https://doi.org/10.1186/s12859-020-03682-4 doi: 10.1186/s12859-020-03682-4
  • This article has been cited by:

    1. Jawdat Alebraheem, Xiangfeng Yang, Predator Interference in a Predator–Prey Model with Mixed Functional and Numerical Responses, 2023, 2023, 2314-4785, 1, 10.1155/2023/4349573
    2. Pan Zheng, Boundedness and global stability in a three-species predator-prey system with prey-taxis, 2023, 0, 1531-3492, 0, 10.3934/dcdsb.2023041
    3. Chuanjia Wan, Pan Zheng, Wenhai Shan, On a quasilinear fully parabolic predator–prey model with indirect pursuit-evasion interaction, 2023, 23, 1424-3199, 10.1007/s00028-023-00931-w
    4. Chuanjia Wan, Pan Zheng, Boundedness and stabilization in an indirect pursuit-evasion model with nonlinear signal-dependent diffusion and sensitivity, 2025, 82, 14681218, 104234, 10.1016/j.nonrwa.2024.104234
    5. N. B. Sharmila, C. Gunasundari, Mohammad Sajid, Mayer Humi, Spatiotemporal Dynamics of a Reaction Diffusive Predator-Prey Model: A Weak Nonlinear Analysis, 2023, 2023, 1687-9651, 1, 10.1155/2023/9190167
    6. Zhoumeng Xie, Yuxiang Li, Global solutions near homogeneous steady states in a fully cross-diffusive predator–prey system with density-dependent motion, 2023, 74, 0044-2275, 10.1007/s00033-023-02127-1
    7. Chuanjia Wan, Pan Zheng, Wenhai Shan, Global stability of a quasilinear predator–prey model with indirect pursuit–evasion interaction, 2024, 17, 1793-5245, 10.1142/S1793524523500766
    8. Pan Zheng, Chuanjia Wan, Global boundedness in a two-competing-predator and one-prey system with indirect prey-taxis, 2025, 0, 1078-0947, 0, 10.3934/dcds.2025001
    9. Minghao Yang, Changcheng Xiang, Yi Yang, Spatiotemporal patterns in a predator–prey model with anti-predation behavior and fear effect, 2025, 2025, 2731-4235, 10.1186/s13662-025-03881-4
    10. Dongze Yan, Global boundedness of Lotka-Volterra competition system with cross-diffusion, 2025, 547, 0022247X, 129346, 10.1016/j.jmaa.2025.129346
  • Reader Comments
  • © 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(3620) PDF downloads(222) Cited by(4)

Figures and Tables

Figures(14)  /  Tables(10)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog