Data processing

Most data were processed using CrystFEL version 0.6.2⁵⁰ directly on the acquired images. Images were pre-selected by our live monitoring software that implemented Cheetah⁵¹ peak finding algorithm 8 for peak detection. Parameters used during the beamtime were: signal to noise ratio = 10, minimum number of pixels per peak = 3 and minimum number of spots 10. For processing optimization, data with minimum SNR = 5 and minimum number of peaks = 20 were used. The final processing with optimized parameters was performed on all images. In this approach, all peaks for indexing were identified using the Zaeferer algorithm as implemented in CrystFel. We choose this approach for simplicity and to avoid unnecessary data duplication.

For 10 runs of the A_2A receptor 6 keV data (corresponding to 5 h of data collection time) we used the EIGER specific implementation of Cheetah⁵¹ developed by Takanori Nakane (https://github.com/biochem-fan/cheetah/tree/eiger) and subsequently used the Cheetah found peaks for indexing.

After indexing and integration using indexamajig, data were scaled without a partiality model in partialator, applying a per pattern resolution cutoff in the process. Data were then converted into mtz files and intensities were converted to structure factor amplitudes by using truncate. All data were kept throughout the process but for the reported statistics and refinement were cut at either CC* falling below 50% or the completeness dropping below 50%.