Multi-scale optimisation vs. genetic algorithms in the gradient separation of diuretics by reversed-phase liquid...

Area of Science:

• Analytical Chemistry
• Chromatography
• Computational Chemistry

Background:

• Multi-linear gradients offer sample separation but are inefficient for complex mixtures.
• Trial-and-error and sequential gradient construction methods are inadequate for highly complex separations.
• Global search methods are necessary for optimizing complex chromatographic gradients.

Purpose of the Study:

• To evaluate and compare the effectiveness of genetic algorithms (GAs) and Multi-scale Optimisation (MSO) for chromatographic separation.
• To determine if GAs can achieve results comparable to MSO in complex sample analysis.
• To assess the impact of penalisation parameters on GA performance.

Main Methods:

• Utilized genetic algorithms (GAs) and Multi-scale Optimisation (MSO) as global search methods for gradient optimization.
• Employed cubic splines and subdivision schemes in MSO to define solvent variation functions.
• Applied restrictions to GAs and MSO, including avoiding long elution times and promoting balanced peak distribution.
• Tested both methods using a C18 column with acetonitrile-water mixtures for separating 14 diuretics and probenecid.

Main Results:

• Both GAs and MSO successfully achieved satisfactory baseline resolution for the complex sample mixture.
• The analysis time for the optimized gradients was approximately 15-16 minutes.
• Genetic algorithms provided results equivalent to those from MSO when penalisation parameters were incorporated into the cost function.

Conclusions:

• Genetic algorithms are a viable and effective alternative to MSO for optimizing complex chromatographic separations.
• The inclusion of penalisation parameters in the cost function is crucial for achieving optimal results with GAs.
• Both GAs and MSO demonstrate the potential to significantly improve separation efficiency for challenging analytical tasks.