FELIX: an algorithm for indexing multiple crystallites in X-ray free-electron laser snapshot diffraction images.
Kenneth R BeyerleinThomas A WhiteOleksandr YefanovCornelius GatiIvan G KazantsevNicolai Fog-Gade NielsenPeter M LarsenHenry N ChapmanSøren SchmidtPublished in: Journal of applied crystallography (2017)
A novel algorithm for indexing multiple crystals in snapshot X-ray diffraction images, especially suited for serial crystallography data, is presented. The algorithm, FELIX, utilizes a generalized parametrization of the Rodrigues-Frank space, in which all crystal systems can be represented without singularities. The new algorithm is shown to be capable of indexing more than ten crystals per image in simulations of cubic, tetragonal and monoclinic crystal diffraction patterns. It is also used to index an experimental serial crystallography dataset from lysozyme microcrystals. The increased number of indexed crystals is shown to result in a better signal-to-noise ratio, and fewer images are needed to achieve the same data quality as when indexing one crystal per image. The relative orientations between the multiple crystals indexed in an image show a slight tendency of the lysozme microcrystals to adhere on ([Formula: see text]10) facets.
Keyphrases
- deep learning
- electron microscopy
- convolutional neural network
- artificial intelligence
- machine learning
- room temperature
- big data
- high resolution
- electronic health record
- crystal structure
- magnetic resonance imaging
- solid state
- smoking cessation
- computed tomography
- data analysis
- mass spectrometry
- quality improvement
- monte carlo
- neural network
- solar cells