Biologically driven neural platform invoking parallel electrophoretic separation and urinary metabolite screening.

Area of Science:

• Analytical Chemistry
• Computational Biology
• Biophysics

Background:

• Electrophoretic separation is crucial for analyzing complex biological samples.
• Multiplexed capillary electrophoresis offers high-throughput analysis but requires complex optimization.
• Existing optimization methods may lack efficiency for multi-parameter systems.

Purpose of the Study:

• To develop and validate a computational framework for optimizing parallel electrophoretic separations.
• To implement a particle swarm optimization (PSO) algorithm on a neural network platform.
• To enhance the separation of biological macromolecules and model urinary metabolites.

Main Methods:

• Utilized a neural network platform integrated with a particle swarm optimization (PSO) algorithm.
• Applied the framework to optimize a 24-capillary electrophoresis system with UV detection.
• Examined the separation of rabbit metallothioneins and toluene urinary metabolites.

Main Results:

• The PSO-optimized neural network demonstrated superior performance over standard backpropagation methods.
• Achieved high accuracy in predicting experimental responses and minimizing training errors.
• Successfully realized full separation of selected analytes under optimal conditions.

Conclusions:

• The proposed computational framework effectively aids in the optimization of parallel electrophoretic separations.
• Metaheuristic computational tools, like PSO, are valuable for chemical separation and screening.
• Provided adaptable pseudo-code for broader implementation of the PSO algorithm.