A statistical approach to detect protein complexes at X-ray free electron laser facilities

Abstract The Flash X-ray Imaging (FXI) technique, under development at X-ray free electron lasers (XFEL), aims to achieve structure determination based on diffraction from individual macromolecular complexes. We report an FXI study on the first protein complex—RNA polymerase II—ever injected at an XFEL. A successful 3D reconstruction requires a high number of observations of the sample in various orientations. The measured diffraction signal for many shots can be comparable to background. Here we present a robust and highly sensitive hit-identification method based on automated modeling of beamline background through photon statistics. It can operate at controlled false positive hit-rate of 3 × 10−5. We demonstrate its power in determining particle hits and validate our findings against an independent hit-identification approach based on ion time-of-flight spectra. We also validate the advantages of our method over simpler hit-identification schemes via tests on other samples and using computer simulations, showing a doubled hit-identification power.