A correlation between crystallographic computing and artificial intelligence research

Machine learning in crystallography and structural science

An overview of the virtual collection on machine learning (ML) in crystallography and structural science, as represented in Acta Crystallographica Sections A, B and D, IUCrJ and Journal of Synchrotron Radiation, is presented. Some terms and concepts related to artificial intelligence and machine learning are briefly introduced and described, and a short history of ML in structural science as it appeared in these IUCr journals is given to whet the appetite for the rest of the collection.

Prediction of models for ordered solvent in macromolecular structures by a classifier based upon resolution-independent projections of local feature data

Current software tools for the automated building of models for macromolecular X-ray crystal structures are capable of assembling high-quality models for ordered macromolecule and small-molecule scattering components with minimal or no user supervision. Many of these tools also incorporate robust functionality for modelling the ordered water molecules that are found in nearly all macromolecular crystal structures. However, no current tools focus on differentiating these ubiquitous water molecules from other frequently occurring multi-atom solvent species, such as sulfate, or the automated building of models for such species. PeakProbe has been developed specifically to address the need for such a tool. PeakProbe predicts likely solvent models for a given point (termed a `peak') in a structure based on analysis (`probing') of its local electron density and chemical environment. PeakProbe maps a total of 19 resolution-dependent features associated with electron density and two associated with the local chemical environment to a two-dimensional score space that is independent of resolution. Peaks are classified based on the relative frequencies with which four different classes of solvent (including water) are observed within a given region of this score space as determined by large-scale sampling of solvent models in the Protein Data Bank. Designed to classify peaks generated from difference density maxima, PeakProbe also incorporates functionality for identifying peaks associated with model errors or clusters of peaks likely to correspond to multi-atom solvent, and for the validation of existing solvent models using solvent-omit electron-density maps. When tasked with classifying peaks into one of four distinct solvent classes, PeakProbe achieves greater than 99% accuracy for both peaks derived directly from the atomic coordinates of existing solvent models and those based on difference density maxima. While the program is still under development, a fully functional version is publicly available. PeakProbe makes extensive use of cctbx libraries, and requires a PHENIX licence and an up-to-date phenix.python environment for execution.

Determining relevant features to recognize electron density patterns in x-ray protein crystallography

High-throughput computational methods in X-ray protein crystallography are indispensable to meet the goals of structural genomics. In particular, automated interpretation of electron density maps, especially those at mediocre resolution, can significantly speed up the protein structure determination process. TEXTAL(TM) is a software application that uses pattern recognition, case-based reasoning and nearest neighbor learning to produce reasonably refined molecular models, even with average quality data. In this work, we discuss a key issue to enable fast and accurate interpretation of typically noisy electron density data: what features should be used to characterize the density patterns, and how relevant are they? We discuss the challenges of constructing features in this domain, and describe SLIDER, an algorithm to determine the weights of these features. SLIDER searches a space of weights using ranking of matching patterns (relative to mismatching ones) as its evaluation function. Exhaustive search being intractable, SLIDER adopts a greedy approach that judiciously restricts the search space only to weight values that cause the ranking of good matches to change. We show that SLIDER contributes significantly in finding the similarity between density patterns, and discuss the sensitivity of feature relevance to the underlying similarity metric.