Nonlinear Dose-Response Modeling of High-Throughput Screening Data Using an Evolutionary Algorithm.

Area of Science:

• Pharmacology and Toxicology
• Computational Biology
• Biostatistics

Background:

• Nonlinear dose-response relationships are prevalent in biological systems under various stresses.
• Accurate modeling is essential for toxicity testing and chemical screening.
• Traditional nonlinear least squares (NLS) methods often fail due to sensitivity to initial parameter values.

Purpose of the Study:

• To develop a robust and versatile method for nonlinear dose-response modeling.
• To address the limitations of traditional fitting methods in computational toxicology.
• To enable automated model selection for diverse dose-response data.

Main Methods:

• Implementation of an evolutionary algorithm (EA) for dose-response curve fitting.
• Application of the EA to a range of common nonlinear models (e.g., exponential, logistic).
• Development of EA's capability for automated model selection without prior assumptions.

Main Results:

• The proposed EA method demonstrates stable and robust fitting of nonlinear dose-response models.
• The EA overcomes the convergence failures and initial value sensitivity associated with NLS.
• The method successfully fits common models and selects the best-fit model from a set of candidates.

Conclusions:

• Evolutionary algorithms provide a superior approach for nonlinear dose-response modeling compared to traditional methods.
• This EA-based approach enhances the reliability and efficiency of toxicity testing and chemical screening.
• The method's ability to fit diverse models and perform model selection is particularly valuable for high-throughput applications.