A Generalized Transform Framework for a Nonlinear Model of Cancer Dynamics

Technical Summary: A Generalized Transform Framework for a Nonlinear Model of Cancer Dynamics

Problem Statement
The paper addresses the challenge of constructing stable, semi-analytical approximations for nonlinear differential models of tumor growth, specifically the logistic–Allee model. Classical growth laws (exponential, logistic, Gompertz) often lack the flexibility to capture density-dependent growth limitations and variable time-scale effects observed in experimental data. While fractional calculus has been applied to such dynamics, standard numerical solutions can be computationally intensive and lack explicit analytical structure. The authors aim to develop a framework that yields compact, semi-analytical representations that are stable, admissible (real-valued over the data range), and comparable in accuracy to standard numerical solvers.

Methodology
The authors propose a "Generalized Laplace–Adomian–Chebyshev–Padé" (GLAP) framework, extended by a $\mu$-scaled generalized transform. The methodology proceeds through the following stages:

1. Model Formulation: The study focuses on a fractional-order extension of the logistic–Allee tumor-growth model:
   $$G^\alpha_T N = aN \left(1 - \frac{N}{b}\right) \left(1 - \frac{c+d}{N+d}\right), \quad N(0) = \eta$$
   where $G^\alpha_T$ is a generalized fractional derivative defined via a positive continuous kernel $T(t, \alpha)$. The model is algebraically reformulated to isolate nonlinear operators ($N^2$ and $(N+d)^{-1}$) suitable for decomposition.

2. Generalized Transform and Decomposition:

   • The authors utilize a generalized Laplace transform $L^\alpha_T$ and the Adomian Decomposition Method (ADM).
   • The nonlinear terms are decomposed into Adomian polynomials ($A_n$ for the rational term and $B_n$ for the quadratic term), allowing for a recursive iterative scheme.
   • A first-order semi-analytical approximation $S_1 = N_0 + N_1$ is derived, where $N_0$ is the zeroth component and $N_1$ is the first correction.

3. Optimization of the Kernel:

   • Instead of prescribing the kernel $T(t, \alpha)$ a priori, it is treated as an optimization-controlled deformation. The authors employ a parametric family: $T(t, \alpha) = \sigma^{1-\alpha} (1 + c(1-\alpha)t)^p$.
   • Parameters $(\alpha, c, \sigma, p)$ are optimized to minimize the error between the approximation $S_1$ and experimental tumor-growth data, subject to "admissibility" constraints. These constraints ensure the transformed variable remains real-valued and that logarithmic terms in the solution do not encounter singularities within the data interval.

4. $\mu$-Scaled Generalized Transform:

   • To further enhance flexibility, a $\mu$-scaled transform is introduced, defining a scaled variable $y_\mu(t) = \mu x(t/\mu)$. This introduces an additional degree of freedom ($\mu$) that affects the domain of the semi-analytical approximation without altering the underlying biological model.
   • The optimization is repeated for the $\mu$-scaled formulation to find parameters that yield the lowest error while maintaining admissibility.

5. Rational Reconstruction:

   • The raw first-order approximation $S_1$ is refined using Chebyshev–Padé rational reconstruction. This step converts the truncated series into a rational function $R_{1,2}$, providing a compact representation that extends the validity of the approximation.

Key Contributions

• Framework Development: The paper constructs a unified GLAP scheme combining generalized Laplace transforms, Adomian decomposition, and Chebyshev–Padé approximants for nonlinear tumor dynamics.
• Admissibility and Stability: The authors establish rigorous admissibility conditions for the transformed approximation. They prove the existence of an admissible minimizer for the transform parameters and characterize the admissible region of the parametric kernel family.
• Fixed-Point Consistency: The work proves that the first-order truncation $S_1$ corresponds to the first nontrivial term of the Adomian series associated with a unique fixed point of the transformed Volterra integral equation. A first-order truncation error estimate is derived.
• Operational Properties: The paper derives inverse formulas, residue inversion formulas, and convolution identities for both the generalized and $\mu$-scaled transforms, providing the theoretical basis for the method.
• Optimization Strategy: A multistage global-local optimization strategy is employed to identify transform parameters that ensure the solution remains real-valued and stable over the experimental data range, avoiding the domain boundaries that often restrict classical formulations.

Results
The proposed framework was tested against experimental tumor-growth data (from Paixão et al.) and compared with standard numerical reference solutions (MATLAB's ode45 and fixed-step RK4).

• Accuracy: The Chebyshev–Padé reconstruction of the first-order approximation yielded Root Mean Square Errors (RMSE) and Maximum Absolute Errors (MAX) comparable to the standard numerical solvers. For instance, the Chebyshev–Padé method achieved an RMSE of ~102.5, while the numerical references were ~102.9.
• Admissibility: Unlike the classical formulation, where the first correction term $N_1$ reaches the boundary of its real admissible domain before covering the full data range, the generalized transform framework successfully maintained a stable, real-valued representation over the entire interval.
• Compactness: The final result is a compact, explicit semi-analytical representation (a rational function of the transformed variable) rather than a purely numerical output.

Significance and Claims
The paper modestly claims that its primary contribution is methodological rather than biological. It does not assert that the generalized transform framework offers superior biological insights over existing models. Instead, it demonstrates that this framework can generate admissible semi-analytical representations for complex nonlinear differential equations where standard approaches might fail due to domain restrictions or lack of explicit structure.

The authors emphasize that the method successfully bridges the gap between semi-analytical tractability and numerical accuracy. By proving the existence of admissible minimizers and establishing fixed-point consistency, the work provides a rigorous theoretical foundation for using generalized transforms in biological modeling. The significance lies in the ability to produce stable, low-order approximations that are computationally efficient and mathematically robust, suitable for further analysis such as parameter sensitivity or identifiability studies in future work.