Improved methods for serial femtosecond crystallography reduction

X-ray free electron lasers (XFELs) are a young technology which alleviates a number of issues, including radiation damage, which remain problematic in conventional macromolecular X-ray crystallography. As it turned out, existing methods had been optimised for synchrotron-based rotation data and coped poorly with snapshot XFEL diffraction images. In the first four results chapters, I present algorithms developed in my efforts to contribute to improved data reduction for XFEL analysis. In order of execution, these are: detector geometry refinement, indexing, initial orientation matrix refinement and post-refinement. Geometry refinement has been a field plagued with the problem of correlated parameters. I introduce the "slip-and-slide" algorithm to solve this by reparameterising detector movements based on their effect on diffraction. I developed an indexing algorithm, TakeTwo, to reduce diffraction pattern wastage by maximising information content on a single image. Initial orientation matrix refinement aims to reduce, without any prior reference, errors of a crystal model from its own information content. When an approximate solution is available, one may implement post-refinement, which, in combination with a model describing the partial illumination of each reflection, drives the model parameters towards a self-consistent data set. In the final results chapter, I present an example of how the combination of all of these algorithms can push the information quality to new levels, illustrated by merging XFEL diffraction collected from scarce quantities of bovine enterovirus type 2 crystals and demonstrating the power of the amplitudes to drive phase extension. I end the thesis with a discussion of the realistic impact of these algorithms, the directions that methods developer communities could take from here onwards, and my personal scientific direction for the future.