VYUŽITÍ MODERNÍCH VÝPOČTOVÝCH METOD A NEURONOVÝCH SÍTÍ K OPTIMALIZACI METODY PRO LC/MS/MS ANALÝZU VOJENSKY VÝZNAMNÝCH ORGANOFOSFOROVÝCH LÁTEK

Significance of the topic

Organophosphorus nerve agents represent some of the most lethal synthetic chemicals ever created. Their rapid inhibition of acetylcholinesterase poses severe threats to human health and security, as demonstrated by recent incidents involving sarin, VX and novichok agents. Analytical methods capable of sensitive, rapid detection in varied matrices are critical for forensic investigations, treaty verification and protective measures.

Goals and Overview of the Study

This study combined density functional theory (DFT)-derived molecular descriptors with neural-network modeling to optimize LC/MS/MS conditions for three military-relevant organophosphorus compounds (A-234, VX, R-33). The objective was to identify chromatographic and ionization settings that maximize precursor-ion signal intensity and to verify whether predictive models trained on one analyte can interpolate optimal conditions for structurally similar agents.

Methodology and Instrumentation

• Analytes: A-234, VX and R-33 prepared at 40 µM in aqueous solution.
• DFT descriptors: HOMO/LUMO energies, dipole moment, molecular entropy.
• Chromatography: Thermo Ultimate 3000 HPLC with ACCUCORE C18 (2.1×150 mm, 2.6 µm), gradient of water/acetonitrile, flow 100–300 µl·min–1, sheath gas flow varied.
• Mass spectrometry: Thermo LTQ XL linear ion trap with ESI source; SIM and MS/MS modes for precursor-ion monitoring and fragment identification.
• Software: Spartan 20 for DFT, MATLAB Deep Learning Toolbox for neural network design, XCalibur/Chromeleon for instrument control, Mass Frontier for fragmentation analysis.
• Neural network: seven inputs (three LC-MS parameters B, Q, E plus four DFT descriptors), one hidden layer with 50 neurons (tanh activation), Bayesian-regulated backpropagation, trained on 396 experimental runs with 84 withheld for validation.

Main Results and Discussion

• Optimal LC and ESI settings differed for each analyte; e.g. A-234: 33 % acetonitrile, 105 µl·min–1 flow, 28 arb sheath gas.
• Addition of 0.1 % formic acid increased signal by up to 36 %; acetic acid provided similar gains; trifluoroacetic acid suppressed response.
• Fragmentation pathways for A-234 were confirmed via Mass Frontier, matching literature assignments.
• Sensitivity improvements: signal enhancement of 1080 %, limits of detection in the low pg·µl–1 or tens of fg·µl–1 range, S/N ≥ 5 at sub-pg levels.
• A-234 hydrolysis in aqueous mobile phase exhibited first-order kinetics (t½ ≈ 4 h 49 min), negligible during typical 30 min analysis time.

Benefits and Practical Applications of the Method

The integrated modeling approach accelerates method development, enabling laboratories (mobile or fixed) to achieve ultratrace detection without extensive empirical trials. Enhanced sensitivity and robustness support forensic and environmental screening of nerve agent residues and their degradation products in complex matrices.

Future Trends and Opportunities

Advances in quantum-chemical descriptor generation and machine learning promise further reduction in experimental iterations. Extending models to new compounds, coupling with microfluidic sample prep, and real-time adaptive control of LC-MS parameters could revolutionize rapid threat detection and broaden applications in pharmaceutical, food and environmental analytics.

Conclusion

This work demonstrates that combining DFT-based molecular descriptors with neural networks can reliably predict optimal LC/MS/MS conditions for organophosphorus nerve agents, improving signal intensity by over tenfold and achieving detection limits in the fg·µl–1 range. The strategy streamlines method development, offering a template for high-throughput analytical workflows in security and beyond.

Reference

• Bajgar J. Adv. Clin. Chem. 38, 151 (2004).
• Lüllmann H., Mohr K., Wehling M. Farmakologie a toxikologie. Grada, Praha 2002.
• Středa L. Fenomén jménem Novičok. Tribun EU, Praha 2022.
• UN General Assembly Security Council: Report A/67/997–S/2013/553 (2013).
• OPCW: Report of the OPCW fact-finding mission on alleged use of chemical weapons in Syria S/1636/2018 (2018).
• OPCW: Decision EC-84/DEC.8 on chemical weapons incident in Kuala Lumpur (2017).
• Stone R. Science 359, 1314 (2018).
• Vale J., Marrs A. T. C., Manynard R. L. Clin. Toxicol. 56, 1093 (2018).
• Nepovidomova E., Kuča K. Arch. Toxicol. 93, 11 (2019).
• Halámek E., Kobliha Z. Chem. Listy 105, 323 (2011).
• Hoenig S. Compendium of Chemical Warfare Agents. Springer 2007.
• Ellison D. H. Handbook of chemical and biological warfare agents, 2 ed. CRC 2007.
• Mirzayanov V. S. Russian Chemical Weapons Program. Outskirts Press 2009.
• Carlsen L. Mol. Inf. 38, 8 (2019).
• Husaini M. A. B. Int. J. Recent Technol. Engineering 8, 3706 (2019).
• Üstün E. et al. J. Organomet. Chem. 815, 16 (2016).
• Pach C. et al. Comput. Ind. 65, 706 (2014).
• Atilgan A. et al. J. Mol. Struct. 1161, 55 (2018).
• Bhakhoa H. et al. R. Soc. Open Sci. 6, 2054 (2019).
• Khalfa A. et al. J. Phys. Chem. A 119, 10527 (2015).
• Li X. et al. J. Phys. Chem. A 113, 10335 (2015).
• Oudejans L. EPA/600/S-19/074 (2019).
• Lee J. Y., Lee Y. H. J. Anal. Chem. 69, 909 (2014).
• Tsuchihashi H. et al. J. Anal. Toxicol. 22, 383 (1998).
• Katagi M. et al. J. Chromatogr. B 698, 81 (1997).
• Bryant P. J. R. et al. J. Chem. Soc. 312, 1553 (1960).
• Sega G. A. et al. J. Chromatogr. A 790, 143 (1997).
• Miki A. et al. J. Anal. Toxicol. 32, 86 (1999).
• Hook G. L. et al. J. Chromatogr. A 992, 1 (2003).
• Mirbabaei F. et al. Anal. Bioanal. Chem. 414, 3429 (2022).
• Hebb D. O. The organization of behavior: a neuropsychological theory, 2 ed. Wiley 1957.
• Gasteiger J., Zupan J. Angew. Chem. Int. Ed. Eng. 32, 503 (1993).
• Zupan J., Rius F. X. Anal. Chim. Acta 239, 311 (1990).
• Röse P., Gasteiger J. Anal. Chim. Acta 235, 163 (1990).
• Otto M. et al. Software Dev. Chem. 4, 377 (1989).
• Kulichenko M. et al. J. Phys. Chem. Lett. 12, 6227 (2021).