RESUMEN
High-intensity femtosecond pulses from an X-ray free-electron laser enable pump-probe experiments for the investigation of electronic and nuclear changes during light-induced reactions. On timescales ranging from femtoseconds to milliseconds and for a variety of biological systems, time-resolved serial femtosecond crystallography (TR-SFX) has provided detailed structural data for light-induced isomerization, breakage or formation of chemical bonds and electron transfer1,2. However, all ultrafast TR-SFX studies to date have employed such high pump laser energies that nominally several photons were absorbed per chromophore3-17. As multiphoton absorption may force the protein response into non-physiological pathways, it is of great concern18,19 whether this experimental approach20 allows valid conclusions to be drawn vis-à-vis biologically relevant single-photon-induced reactions18,19. Here we describe ultrafast pump-probe SFX experiments on the photodissociation of carboxymyoglobin, showing that different pump laser fluences yield markedly different results. In particular, the dynamics of structural changes and observed indicators of the mechanistically important coherent oscillations of the Fe-CO bond distance (predicted by recent quantum wavepacket dynamics21) are seen to depend strongly on pump laser energy, in line with quantum chemical analysis. Our results confirm both the feasibility and necessity of performing ultrafast TR-SFX pump-probe experiments in the linear photoexcitation regime. We consider this to be a starting point for reassessing both the design and the interpretation of ultrafast TR-SFX pump-probe experiments20 such that mechanistically relevant insight emerges.
Asunto(s)
Artefactos , Rayos Láser , Mioglobina , Cristalografía/instrumentación , Cristalografía/métodos , Electrones , Mioglobina/química , Mioglobina/metabolismo , Mioglobina/efectos de la radiación , Fotones , Conformación Proteica/efectos de la radiación , Teoría Cuántica , Rayos XRESUMEN
The use of artificial intelligence to process diffraction images is challenged by the need to assemble large and precisely designed training data sets. To address this, a codebase called Resonet was developed for synthesizing diffraction data and training residual neural networks on these data. Here, two per-pattern capabilities of Resonet are demonstrated: (i) interpretation of crystal resolution and (ii) identification of overlapping lattices. Resonet was tested across a compilation of diffraction images from synchrotron experiments and X-ray free-electron laser experiments. Crucially, these models readily execute on graphics processing units and can thus significantly outperform conventional algorithms. While Resonet is currently utilized to provide real-time feedback for macromolecular crystallography users at the Stanford Synchrotron Radiation Lightsource, its simple Python-based interface makes it easy to embed in other processing frameworks. This work highlights the utility of physics-based simulation for training deep neural networks and lays the groundwork for the development of additional models to enhance diffraction collection and analysis.