Structure-response relationship in electrospray ionization-mass spectrometry of sartans by artificial neural networks
Само за регистроване кориснике
2016
Чланак у часопису (Објављена верзија)
Метаподаци
Приказ свих података о документуАпстракт
Quantitative structure-property relationship (QSPR) methods are based on the hypothesis that changes in the molecular structure are reflected in changes in the observed property of the molecule. Artificial neural network is a technique of data analysis, which sets out to emulate the human brain's way of working. For the first time a quantitative structure-response relationship in electrospray ionization mass spectrometry (ESI-MS) by means of artificial neural networks (ANN) on the group of angiotensin II receptor antagonists - sartans has been established. The investigated descriptors correspond to different properties of the analytes: polarity (logP), ionizability (pKa), surface area (solvent excluded volume) and number of proton acceptors. The influence of the instrumental parameters: methanol content in mobile phase, mobile phase pH and flow rate was also examined. Best performance showed a multilayer perceptron network with the architecture 6-3-3-1, trained with backpropagation alg...orithm. It showed high prediction ability on the previously unseen (test) data set with a coefficient of determination of 0.994. High prediction ability of the model would enable prediction of ESI-MS responsiveness under different conditions. This is particularly important in the method development phase. Also, prediction of responsiveness can be important in case of gradient-elution LC-MS and LC-MS/MS methods in which instrumental conditions are varied during time. Polarity, chargeability and surface area all appeared to be crucial for electrospray ionization whereby signal intensity appeared to be the result of a simultaneous influence of the molecular descriptors and their interactions. Percentage of organic phase in the mobile phase showed a positive, while flow rate showed a negative impact on signal intensity.
Извор:
Journal of Chromatography A, 2016, 1438, 123-132Издавач:
- Elsevier Science BV, Amsterdam
Финансирање / пројекти:
- Синтеза, квантитативни однос између структуре и дејства, физичко-хемијска карактеризација и анализа фармаколошки активних супстанци (RS-MESTD-Basic Research (BR or ON)-172033)
- German Academic Exchange Service (DAAD)
DOI: 10.1016/j.chroma.2016.02.021
ISSN: 0021-9673
PubMed: 26884139
WoS: 000371941500013
Scopus: 2-s2.0-84959177424
Институција/група
PharmacyTY - JOUR AU - Golubović, Jelena AU - Birkemeyer, Claudia AU - Protić, Ana AU - Otašević, Biljana AU - Zečević, Mira PY - 2016 UR - https://farfar.pharmacy.bg.ac.rs/handle/123456789/2541 AB - Quantitative structure-property relationship (QSPR) methods are based on the hypothesis that changes in the molecular structure are reflected in changes in the observed property of the molecule. Artificial neural network is a technique of data analysis, which sets out to emulate the human brain's way of working. For the first time a quantitative structure-response relationship in electrospray ionization mass spectrometry (ESI-MS) by means of artificial neural networks (ANN) on the group of angiotensin II receptor antagonists - sartans has been established. The investigated descriptors correspond to different properties of the analytes: polarity (logP), ionizability (pKa), surface area (solvent excluded volume) and number of proton acceptors. The influence of the instrumental parameters: methanol content in mobile phase, mobile phase pH and flow rate was also examined. Best performance showed a multilayer perceptron network with the architecture 6-3-3-1, trained with backpropagation algorithm. It showed high prediction ability on the previously unseen (test) data set with a coefficient of determination of 0.994. High prediction ability of the model would enable prediction of ESI-MS responsiveness under different conditions. This is particularly important in the method development phase. Also, prediction of responsiveness can be important in case of gradient-elution LC-MS and LC-MS/MS methods in which instrumental conditions are varied during time. Polarity, chargeability and surface area all appeared to be crucial for electrospray ionization whereby signal intensity appeared to be the result of a simultaneous influence of the molecular descriptors and their interactions. Percentage of organic phase in the mobile phase showed a positive, while flow rate showed a negative impact on signal intensity. PB - Elsevier Science BV, Amsterdam T2 - Journal of Chromatography A T1 - Structure-response relationship in electrospray ionization-mass spectrometry of sartans by artificial neural networks VL - 1438 SP - 123 EP - 132 DO - 10.1016/j.chroma.2016.02.021 ER -
@article{ author = "Golubović, Jelena and Birkemeyer, Claudia and Protić, Ana and Otašević, Biljana and Zečević, Mira", year = "2016", abstract = "Quantitative structure-property relationship (QSPR) methods are based on the hypothesis that changes in the molecular structure are reflected in changes in the observed property of the molecule. Artificial neural network is a technique of data analysis, which sets out to emulate the human brain's way of working. For the first time a quantitative structure-response relationship in electrospray ionization mass spectrometry (ESI-MS) by means of artificial neural networks (ANN) on the group of angiotensin II receptor antagonists - sartans has been established. The investigated descriptors correspond to different properties of the analytes: polarity (logP), ionizability (pKa), surface area (solvent excluded volume) and number of proton acceptors. The influence of the instrumental parameters: methanol content in mobile phase, mobile phase pH and flow rate was also examined. Best performance showed a multilayer perceptron network with the architecture 6-3-3-1, trained with backpropagation algorithm. It showed high prediction ability on the previously unseen (test) data set with a coefficient of determination of 0.994. High prediction ability of the model would enable prediction of ESI-MS responsiveness under different conditions. This is particularly important in the method development phase. Also, prediction of responsiveness can be important in case of gradient-elution LC-MS and LC-MS/MS methods in which instrumental conditions are varied during time. Polarity, chargeability and surface area all appeared to be crucial for electrospray ionization whereby signal intensity appeared to be the result of a simultaneous influence of the molecular descriptors and their interactions. Percentage of organic phase in the mobile phase showed a positive, while flow rate showed a negative impact on signal intensity.", publisher = "Elsevier Science BV, Amsterdam", journal = "Journal of Chromatography A", title = "Structure-response relationship in electrospray ionization-mass spectrometry of sartans by artificial neural networks", volume = "1438", pages = "123-132", doi = "10.1016/j.chroma.2016.02.021" }
Golubović, J., Birkemeyer, C., Protić, A., Otašević, B.,& Zečević, M.. (2016). Structure-response relationship in electrospray ionization-mass spectrometry of sartans by artificial neural networks. in Journal of Chromatography A Elsevier Science BV, Amsterdam., 1438, 123-132. https://doi.org/10.1016/j.chroma.2016.02.021
Golubović J, Birkemeyer C, Protić A, Otašević B, Zečević M. Structure-response relationship in electrospray ionization-mass spectrometry of sartans by artificial neural networks. in Journal of Chromatography A. 2016;1438:123-132. doi:10.1016/j.chroma.2016.02.021 .
Golubović, Jelena, Birkemeyer, Claudia, Protić, Ana, Otašević, Biljana, Zečević, Mira, "Structure-response relationship in electrospray ionization-mass spectrometry of sartans by artificial neural networks" in Journal of Chromatography A, 1438 (2016):123-132, https://doi.org/10.1016/j.chroma.2016.02.021 . .