Signal extraction in SWAXS data for the compact X-ray light sources: a machine learning approach.

The development of X-ray free electron lasers has driven significant progress in X-ray science. Given the broad range of their applications, implementing a new generation of this technology at the laboratory scale has been under consideration for several years. This initiative is now under commissioning and construction at Arizona State University, known as the Compact X-ray Light Source (CXLS) and the Compact X-ray Free Electron Laser (CXFEL). Alongside experimental advances in this direction, whether in large or compact X-ray free electron lasers, there is also a growing need for new algorithmic and analytical methods to process the data obtained from such facilities. This work introduces a novel approach for analyzing Small- and Wide-Angle X-ray Scattering (SWAXS) profiles using a data-driven machine learning algorithm. The method is proposed for application to SWAXS datasets collected at both compact and large-scale X-ray facilities. To evaluate the performance of this approach, we analyzed simulated time-resolved SWAXS data from a protein, generated based on the current CXLS experimental parameters, and compared the results with those from the standard singular value decomposition (SVD) technique. Despite the low photon counts in the data, the results demonstrate that our method achieves higher accuracy in extracting structural dynamics information compared to SVD.