BMC Bioinformatics Молекулярная биология

Robust model calibration in large-scale systems biology via an adaptive hybrid algorithm

Polo-Rodríguez, Andrea, R. Penas, David, R. Banga, Julio
Springer Nature · 2026-11-12 · DOI 10.1186/s12859-026-06727-2

Аннотация

Background Parameter estimation is a central challenge in systems biology, particularly for large dynamic models described by nonlinear ordinary differential equations (ODEs). These global optimization problems exhibit landscapes which are topologically heterogeneous, often exhibiting a pathological mixture of stiff, smooth valleys and rugged, noisy plateaus, making single-strategy hybrids ineffective. While methods like enhanced Scatter Search (eSS) represent the current state-of-the-art, their rigid intensification strategies can limit performance in large-scale or ill-conditioned scenarios. Results We introduce eLSHADE+, a novel metaheuristic architecture designed to adapt to these topological challenges. The proposed algorithm builds on the well-established Differential Evolution variant LSHADE by incorporating an adaptive multi-strategy hybridization. Unlike standard memetic algorithms, eLSHADE+ uses three distinct operational modes: (i) gradient-based intensification, (ii) derivative-free search, and (iii) pure global exploration. Adaptively switching the intensification strategy allows the algorithm to capture the complementary strengths of different local optimization paradigms while mitigating their individual limitations, achieving precision in differentiable regions, robustness against numerical noise, and saving computational resources in complicated basins. Additionally, a logarithmic parameter space transformation is incorporated to facilitate search across multi-scale biological constants. Conclusions We evaluated eLSHADE+ using the BioPreDyn benchmark suite, comprising realistic, biologically motivated problems. Within this benchmark suite and under the reported experimental conditions, the proposed multi-strategy approach achieved statistically competitive or superior convergence speed and solution accuracy compared with the methods evaluated. Code and data are available in a publicly accessible repository. These results support eLSHADE+ as a competitive and robust alternative for the model-calibration problems considered.

Читать у издателя →