Vincent J. Fazio

Practical Aspects of the SAMPL Challenge: Providing an Extensive Experimental Data Set for the Modeling Community

To provide an experimental basis for a comprehensive molecular modeling evaluation study, 500 fragments from the Maybridge fragment library were soaked into crystals of bovine pancreatic trypsin and the structures determined by X-ray crystallography. The soaking experiments were performed in both single and pooled aliquots to determine if combination of fragments is an appropriate strategy. A further set of data was obtained from co-crystallizing the pooled fragments with the protein. X-ray diffraction data were collected on approximately 1000 crystals at the Australian Synchrotron, and these data were subsequently processed, and the preliminary analysis was performed with a custom software application (Jigsaw), which combines available software packages for structure solution and analysis.

The DINGO dataset: a comprehensive set of data for the SAMPL challenge

Part of the latest SAMPL challenge was to predict how a small fragment library of 500 commercially available compounds would bind to a protein target. In order to assess the modellers’ work, a reasonably comprehensive set of data was collected using a number of techniques. These included surface plasmon resonance, isothermal titration calorimetry, protein crystallization and protein crystallography. Using these techniques we could determine the kinetics of fragment binding, the energy of binding, how this affects the ability of the target to crystallize, and when the fragment did bind, the pose or orientation of binding. Both the final data set and all of the raw images have been made available to the community for scrutiny and further work. This overview sets out to give the parameters of the experiments done and what might be done differently for future studies.

Evaluating the solution from MrBUMP and BALBES

The automated pipelines for molecular replacement MrBUMP and BALBES are reviewed, with an emphasis on understanding their output. Conclusions are drawn from their performance in extensive trials.

A drunken search in crystallization space.

The REMARK280 field of the Protein Data Bank is the richest open source of successful crystallization information. The REMARK280 field is optional and currently uncurated, so significant effort needs to be applied to extract reliable data. There are well over 15 000 crystallization conditions available commercially from 12 different vendors. After putting the PDB crystallization information and the commercial cocktail data into a consistent format, these data are used to extract information about the overlap between the two sets of crystallization conditions. An estimation is made as to which commercially available conditions are most appropriate for producing well diffracting crystals by looking at which commercial conditions are found unchanged (or almost unchanged) in the PDB. Further analyses include which commercial kits are the most appropriate for shotgun or more traditional approaches to crystallization screening. This analysis suggests that almost 40% of the crystallization conditions found currently in the PDB are identical or very similar to a commercial condition.

DroplIT, an improved image analysis method for droplet identification in high-throughput crystallization trials

The application of robotics to protein crystallization trials has resulted in the production of millions of images. Manual inspection of these images to find crystals and other interesting outcomes is a major rate-limiting step. As a result there has been intense activity in developing automated algorithms to analyse these images. The very first step for most systems that have been described in the literature is to delineate each droplet. Here, a novel approach that reaches over 97% success rate and subsecond processing times is presented. This will form the seed of a new high-throughput system to scrutinize massive crystallization campaigns automatically.

Fragment Screening for the Modelling Community: SPR, ITC, and Crystallography

The SAMPL (Statistical Assessment of the Modelling of Proteins and Ligands) challenge brought together experimentalists and modellers in an effort to improve our understanding of chemical and biochemical systems so better modelling tools can be developed. The most recent challenge, SAMPL3, held at Stanford University in August 2011, was an attempt to improve the methods used to predict how small fragment compounds bind to proteins, and the protein chosen for this test was bovine trypsin. Surface plasmon resonance was used to screen 500 compounds from a Maybridge fragment library and these compounds were subsequently used to soak crystals of trypsin and the best hits were also characterised by isothermal titration calorimetry. We present methods used for the surface plasmon resonance and the isothermal titration calorimetry experiments, as well as the results for these methods and those compounds that were found in the crystal structures.

A universal indicator dye pH assay for crystallization solutions and other high-throughput applications.

In protein crystallization, as well as in many other fields, it is known that the pH at which experiments are performed is often the key factor in the success or failure of the trials. With the trend towards plate-based high-throughput experimental techniques, measuring the pH values of solutions one by one becomes prohibitively time- and reagent-expensive. As part of an HT crystallization facility, a colour-based pH assay that is rapid, uses very little reagent and is suitable for 96-well or higher density plates has been developed.

Crystallization: Digging into the Past to Learn Lessons for the Future

Crystals of biological macromolecules have been observed and grown for well over a century. More effort has been put into biological crystallization in the last few decades due to the importance of X-ray crystal structures, the advent of synchrotron radiation sources, improved computational speed, better software, and the availability of recombinant protein. Here we focus on two important areas of crystal growth: firstly, on techniques for stabilizing the protein sample, and secondly, on strategies and approaches for selecting the crystallization cocktails most suitable for different strategies.

Classification of protein crystallisation images using texture-based statistical features

We report on the classification of protein crystallisation data via decision trees using N-fold cross validation. Protein crystallisation images were obtained over a period of time ranging from days to months. All the images taken of a single experiment are arranged according to the time they were obtained and aligned according to the position of the experimental droplet. Difference between consecutive images in a time course was determined and the background noise was removed using filtering techniques. Image analysis is performed on the area inside the droplet of each difference image to compute statistical texture features. Classification is carried out by testing and training with 1000 sample images and an accuracy rate of ∼75% was achieved.

The

The PDB is currently growing at a rate of about 9000 structures annually, and 90% of these have been determined by X-ray diffraction methods. Each structure is the result of one or more crystals. Not every protein crystallises nor do all crystals diffract well enough; it has been estimated that of every 10 proteins that are purified, four will show some sign of crystallisation and one will crystallise robustly enough to obtain a structure1. Using these numbers, and making some educated guesses (for example, that most proteins are tested in 1000 crystallisation trials1) these 8000 structures represent 80,000 purified proteins, and 80,000,000 crystallisation trials which are set up each year. The cost of consumables, chemicals and direct labour to set up those trials varies, but can be estimated to be