The interpretation of Fourier maps is an
important process in X-ray diffraction analysis. Their quality and resolution
can be limited by a number of factors, including the disorder of atoms in the
crystal and poorly measured diffraction data, which can be caused, for example,
by crystal instability during the experiment, twinning, small crystal size, or
otherwise poor quality of the measured crystals. These imperfections often lead
to blurred or difficult-to-interpret maps, which complicate the determination
of the crystal structure and cause headaches for many crystallographers.
In this work, we present a possible approach that could help interpret low-resolution maps. This approach uses deep learning capabilities to recognize and reconstruct relevant features in maps. In the reconstructed maps, noise is removed and blurred parts of the map, which appear to be meaningless blobs, are interpreted. This approach can be applied at various stages of structural analysis. For example, when interpreting maps after solving a phase problem or interpreting differential Fourier maps. This approach is also applicable to various groups of substances and materials, regardless of whether they are inorganic, organic, or even protein structures, thus opening up new possibilities for streamlining the entire structural analysis process.