Download citation
Acta Cryst. (2014). A70, C1628
Download citation

link to html
We show that suitably chosen machine learning algorithms can be used to predict the "crystallisation propensity" of classes of molecules with a promisingly low error rate, using the Cambridge Structural Database and ZINC database to provide training examples of crystalline and non-crystalline molecules. Supervised learning tasks involve using machine learning algorithms to infer a function from known training data which allows classification of unknown test data. Such algorithms have been successfully used to predict continuous properties of compounds, such as melting point[1] and solubility[2]. Similar methods have also been applied to protein crystallinity predictions based on amino acid sequences[3], but little has previously been done to attempt to classify small organic molecules as crystalline or non-crystalline due to the difficulty in finding descriptors appropriate to the problem. Our approach uses only information about the atomic types and connectivity, leaving aside the confounding effects of solvents and crystallisation conditions. The result is reinforced by a blind microcrystallisation screening of a sample of materials, which confirmed the classification accuracy of the predictive model. An analysis of the most significant descriptors used in the classification is also presented, and we show that significant predictive accuracy can be obtained using relatively few descriptors.

Download citation
Acta Cryst. (2014). A70, C1720
Download citation

link to html
Since the early 70's, multiple wavelength experiments have been used to determine phases of proteins containing anomalous scatterers. The small molecule single crystal beamline, I19,[1] at Diamond Light Source, is designed to carry out single crystal anomalous dispersion studies using tunable wavelength. These experiments can differentiate between oxidation states; discriminate between atoms with near-identical X-ray scattering factors; and solve the phase problem for very low resolution X-ray data. We describe the application of MAD phasing (Karle and Hendrickson [2]) to determine the structure of large `small molecules' where only low-resolution data is available. Initial studies were carried out on a known, (well diffracting) centrosymmetric bromide containing compound. The wavelength dependence of the anomalous signal from the bromide was calculated from fluorescence absorption data in DetOx.[3] Datasets were then collected at 4 wavelengths chosen to maximize differences in the anomalous signal. Using the MAD phasing equations we obtained estimates for the anomalous scattering contribution from all atoms in the structure and a phase difference between that and the normal scattering component. This allowed us to reduce noise in the Patterson map and locate only the heavy atom scatterers. We then use phase estimates from the heavy atom substructure to locate the rest of the atoms. Initial proof of concept experiments will now be extended to larger structures where data is not of sufficient resolution to be solved by direct methods alone.

Download citation
Acta Cryst. (2014). A70, C1728
Download citation

link to html
Standard crystallographic structure refinements employ anisotropic displacement parameters (ADPs) to represent the probability distribution of a scattering atom. Such distributions may be due to thermal motion of the atom and / or a spatial average of multiple discrete atomic positions. An anisotropic description of an atomic distribution requires six parameters, and - in cases where data is limited or poor quality - the optimal values of these parameters may be ill-defined. Application of restraints and constraints can impose some physical and chemical reality on the set of displacement parameters. Examples include those based on the Hirshfeld Rigid Bond Test [1], and more recently SHELXL's RIGU [2]. We have implemented these and other a.d.p. restraints in CRYSTALS [3], for introducing reasonable relationships amongst common arrangements of anisotropic atoms. Use of a priori information in the form of restraints must always be justified, and we present an assessment of the applicability of the new restraints against a large data set of high quality crystal structure determinations.

Download citation
Acta Cryst. (2014). A70, C1750
Download citation

link to html
Approaches to determining the influence of individual measurements on the precision of crystallographic least squares parameters have been known for a long while.[1] Situations in which the precision of a single parameter (or linear combination of parameters) is critical can include: determination of novel bond lengths; refinement of site occupancies in mixed metal or mixed oxidation state systems; determination of the fraction of excited state molecules in a time-resolved pump-probe experiment. Such calculations are easily applicable to point-detector instruments, where individual influential reflections could be remeasured one-by-one. However, on a modern area detector instrument many reflections are measured on one frame and therefore some consideration of the appropriate strategy of reciprocal space scans is permitted to allow a more efficient use of the instrument. The highly influential partial data collection is then feed into an appropriate refinement model. Occupancies in mixed-metal or mixed-oxidation state systems and fractions and positions of excited state molecules during a time-resolved pump-probe experiment can be determined using direct refinement of the perturbation of the structure from the ground state. Re-factoring to modern Fortran of the Crystals software is in progress to allow the implementation of new algorithms such as a difference refinement.[2] We present an analysis of diffractometer strategy selection to prioritize scans which give the best improvement in specific least-squares parameters and a novel algorithm for the refinement of the partial data using crystals.
Follow Acta Cryst. A
Sign up for e-alerts
Follow Acta Cryst. on Twitter
Follow us on facebook
Sign up for RSS feeds