Answers are grounded in this paper and cite their sources.
Structure determination of organic molecular crystals relies heavily on single-crystal X-ray diffraction, yet suitable single crystals are often difficult to obtain. A common alternative uses crystal structure prediction (CSP) to generate candidate structures for subsequent comparison with experimental powder X-ray diffraction (PXRD) patterns. Such trial-and-error searches across multiple space groups incur substantial computational costs. Here, we introduce a structure determination protocol that uses experimental PXRD together with molecular structure as inputs to guide CSP from the outset. A multimodal neural network predicts probabilities for candidate space groups, which are then searched sequentially in descending probability order. Candidate structures are evaluated using PXRD agreement and relative energies, and the search terminates when a candidate satisfies the matching criterion. In a retrospective evaluation of 11 molecular-crystal systems, the protocol returned candidate sets containing structures matching the experimental crystal packing in nine cases. Compared with exhaustive searches across the 22 most common space groups, it reduced the median computational cost of structure determination by 97.9%. We implement this iterative protocol as a fully automated agentic workflow that autonomously coordinates space-group prediction, structure generation, optimization and diffraction matching, enabling crystal structure determination from molecular structure and experimental PXRD.
No takes yet. Share an insight, caveat, or question.
Qiao et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: