Összes szerző

---

Sipos Áron

az alábbi absztraktok szerzői között szerepel:

Sipos Áron
Machine learning analysis of ultrafast fluorescence kinetics of NADH conformations in solutions

Aug 31 - csütörtök

10:30 – 10:45

Bioenergetika és fotobiofizika

E40

Machine learning analysis of ultrafast fluorescence kinetics of NADH conformations in solutions

Áron Sipos, Ferenc Sarlós, Rita Nagypál, László Zimányi and Géza I. Groma

Institute of Biophysics, Biological Research Centre Szeged, Eötvös Loránd Research Network, Szeged, Hungary

The structure of many different enzyme-bound forms of the essential coenzyme nicotinamide adenine dinucleotide is well characterized by X-ray diffraction data. Due to the limitations of this technique in solution, the unbound forms of the molecule need to be characterized by alternative methods, such as time-resolved fluorescence spectroscopy. In NADH the relative position of the nicotinamide and adenine groups has primary impact on the fluorescence kinetics of the excited nicotinamide group. In aqueous solution the molecule exists in an equilibrium of closed and open conformations, while the presence of methanol favors the latter.

The fluorescence kinetics of NADH were measured in water and methanol environments using fluorescence upconversion and time-correlated single photon counting in a large, 50 fs – 10 ns time window at different wavelengths for both environments.

To avoid the uncertainties of exponential fitting, the experimental data were fitted by a quasi-continuous set of time constants, applying regularization terms for favoring sparse solutions, i.e., a minimum number of nonzero amplitudes. For fine tuning the level of sparsity we developed a machine-learning method based on cross-validation and Bayesian optimization. This approach was found to be a powerful method for fluorescence kinetics analysis, avoiding any arbitrary or random parameters.

According to the above analysis the fast (<100 ps) part of the kinetics can be characterized by an unusually complex, three-step vibrational relaxation process. The slow part is well modelled either by the conventional distinct exponential terms or by distributed kinetics, corresponding to an equilibrium of a very high number of conformational states [1], as shown by the improved version of the analysis method.

Acknowledgment

This work has been supported by the National Research, Development and Innovation Office of Hungary; 2018-1.2.1-NKP-2018-00009; ÁS is grateful for the support of National Research, Development and Innovation Office of Hungary PD-121170.

References

[1] Zimányi L, Sipos Á, Sarlós F, Nagypál R, Groma GI (2021) PLoS ONE 16(8): e0255675.