Optimisation of HPLC gradient separations using artificial neural networks (ANNs): application to benzodiazepines in...

Area of Science:

• Analytical Chemistry
• Computational Chemistry

Background:

• Optimizing gradient High-Performance Liquid Chromatography (HPLC) separations is crucial for analyzing complex mixtures.
• Traditional methods for gradient optimization can be time-consuming and labor-intensive.
• Artificial neural networks (ANNs) offer a data-driven approach to complex system modeling.

Purpose of the Study:

• To develop and validate an optimized gradient HPLC method for the separation of nine benzodiazepines using artificial neural networks.
• To assess the predictive accuracy and efficiency of ANNs in chromatographic method development.
• To apply the optimized method for the analysis of benzodiazepines in authentic post-mortem samples.

Main Methods:

• Artificial neural networks (ANNs) were employed in conjunction with an experimental design.
• The ANNs were trained to predict retention times and peak widths for nine benzodiazepines.
• Optimization focused on parameters such as buffer concentration and pH, organic modifier composition, and gradient profile.

Main Results:

• The optimal conditions predicted by the best-performing ANN included specific buffer, methanol, and acetonitrile concentrations and gradient settings.
• The predictive error for retention times and peak widths was less than 5% for six out of nine analytes.
• The optimized method demonstrated limits of detection (LODs) between 0.0057-0.023 microg/mL and recoveries ranging from 58% to 92%.

Conclusions:

• ANNs provide a flexible and convenient approach for optimizing gradient elution separations in HPLC.
• The developed method is effective for the analysis of benzodiazepines in complex matrices like post-mortem samples.
• This study highlights the potential of ANNs to streamline and improve chromatographic method development.