Description of binary system by artificial neural networks

Investor logo

Warning

This publication doesn't include Faculty of Economics and Administration. It includes Faculty of Science. Official publication website can be found on muni.cz.
Authors

TOPINKOVÁ Jana TRNKOVÁ Libuše FARKOVÁ Marta HAVEL Josef

Year of publication 2004
Type Article in Proceedings
Conference Sborník příspěvků IV.pracovní setkání fyzikálních chemiků a elektrochemiků
MU Faculty or unit

Faculty of Science

Citation
Field Electrochemistry
Keywords Binary system; adenine; cytosine; UV/Vis spectrophotometry; artificial neural networks (ANNs);
Description There is a growing interest in using artificial neural networks (ANN), not only in chemistry, but also in other research areas. Some recent applications of ANN have been in multicomponent analysis, kinetics, optimization of processes in HPLC and CZE, in chemical equilibria computations, and in spectrophotometry. Adenine (Ade) and cytosine (Cyt) are of great interest for several reasons; in particular, they are the important components of nucleic acids and coenzymes, playing an important role in many biological processes. Generally, the concentrations of individual nucleic bases are usually determined by UV/VIS spectrophotometry. The absorption maxima of these compounds are at 263nm (Ade) and 271 nm (Cyt) at pH 4.7. In this case of fully overlapped spectra, multicomponent analysis can solve the problem. The aim of this work was to explain and show practical possibilities of ANN for an advanced laboratory course in physical chemistry. In this work, students (1) to record separately spectra of Ade and Cyt; (2) to prepare mixtures of Ade and Cyt according to experimental design (23 factorial design) in a concentration range from 0.1 to 0.5 mM, and in 0.5 M acetate buffer (pH 4.7 and ionic strength 0.5); (3) to measure non-smoothed spectra from 240 to 290 nm; (4) to prepare input data (absorbances) and choose the required output data (Ade and Cyt concentrations); (5) to search for the optimal ANN structure (number of input, hidden, and output neurons) and optimal number of epochs for the adaptation (training) process; (6) to use suitable test data (not included in the training set) for verification and testing; (7) to apply the optimal ANN for analysis of unknown samples. The quality of the prediction for the test set was evaluated on the basis of the average root mean square error for prediction (RMSEP) calculated from given and predicted values of Ade and Cyt concentrations. The data were processed on a Pentium PC computer using TRAJAN Release 3.0D software package from TRAJAN Software LTD (Durham, U. K).
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.