engleski

Artificial Neural Networks in UV-VIS Spectral Analysis of Ribonucleotides

The ability of ANNs to learn process inherent features and to detect characteristic patterns with a little apriori knowledge, makes ANNs powerful and flexible tools in the various chemometrics studies. The precise quantitative discrimination of quite similar ribonucleotide spectrophotometric data in UV range (210-300 nm) is necessary in a wide range of applications (e. g., the flavor enhancers production). Several types of ANNs, such as multilayer perceptrons, generalized feedforward networks and modular feedforward networks, with different topologies, were applied for the quantitative identification of the concentration of 5'-CMP, 5'-GMP, 5'-UMP, 5'-AMP, 2'- and 3'-GMP, 2'- and 3'-UMP. The achieved results in resolving spectral overlapping of pure components is the first, essential step for the extension of this artificial intelligence methodology to the complex mixtures resolution. This instrumental approach will avoid time-consuming cleanup procedures or separation steps. As because the strength of ANNs lies in their ability to generalize complex hidden relationship, careful cross validation supervision was performed.

neural networks; spectroscopy; RNA; classification

nije evidentirano