Related Experiment Videos
Development of migration models for acids in capillary electrophoresis using heuristic and radial basis function
Chunxia Xue1, Xiaojun Yao, Huanxiang Liu
1Department of Chemistry, Lanzhou University, Lanzhou, China.
Electrophoresis
|April 27, 2005
Summary
A quantitative structure-mobility relationship was developed for acids using computational models. Radial basis function neural networks provided more accurate mobility predictions than heuristic linear models.
Area of Science:
- Analytical Chemistry
- Computational Chemistry
Background:
- Capillary electrophoresis (CE) is a powerful separation technique.
- Predicting analyte mobility is crucial for method development and optimization in CE.
Purpose of the Study:
- To develop quantitative structure-mobility relationships (QSMR) for predicting the absolute mobilities of organic and inorganic acids in CE.
- To compare the performance of linear (heuristic method) and nonlinear (radial basis function neural networks) models for mobility prediction.
Main Methods:
- Developed a QSMR model using descriptors calculated solely from molecular structure.
- Employed the heuristic method (HM) for linear model construction.
- Utilized radial basis function neural networks (RBFNN) for nonlinear model construction.
- Validated models using training, test, and whole datasets.
Main Results:
- Both HM and RBFNN models showed good agreement with experimental mobilities.
- RBFNN models achieved lower root-mean-square (RMS) errors compared to HM models across all datasets.
- HM model provided insights into factors influencing acid mobilities.
Conclusions:
- QSMR models can accurately predict acid mobilities in capillary electrophoresis.
- Nonlinear RBFNN models offer superior predictive performance over linear HM models for this application.
- Structure-based prediction of electrophoretic mobility is a viable approach for chemical analysis.