Spectrophotometric analysis of nucleic acid bases in binary and ternary mixtures by partial least squares and artificial neural networks
| Authors | |
|---|---|
| Year of publication | 2006 |
| Type | Article in Proceedings |
| Conference | 3rd International Symposium on Computer Applications and Chemometrics in Analytical Chemistry SCAC 2006 |
| MU Faculty or unit | |
| Citation | |
| Field | Analytic chemistry |
| Keywords | artificial neural networks; partial least squares; adenine; cytosine; thymine; UV-Vis spectrophotometry; |
| Description | The contribution shows a comparative study of the use of partial least squares (PLS) and artificial neural networks (ANNs) to analyze nucleic acid bases (adenine - A, cytosine - C, thymine -T) in mixtures by UV-Vis spectrophotometry. Multivariate calibration based on a suitable experimental design (ED) and soft modeling for the quantitative analysis of fully overlapped ultraviolet spectra of A+C binary and/or A+C+T ternary systems were employed. The optimal ANN architecture, enabling to model the system, was established by means of TRAJAN program. Using the multivariate statistical method SIMCA the linearity of the PLS model for dilute solutions containing two or three bases in acetate buffer (pH 4.7) was proved. A combination of the chemometric methods and derivative spectrophotometry was applied. The ability of ANN and PLS to model binary and ternary systems was evaluated on the basis of the root means square error for prediction (RMSEP) and the agreement factor (AF). For the determination of A, C, and T concentrations in binary and/or ternary mixtures, the relative standard deviation was less than 1 %. High predictive potentials of both chemometric methods have been demonstrated. |
| Related projects: |
|