Purpose of the fragment library

The fragment library is intended for drug discovery by fragment screening. It is suitable for crystallographic, NMR or SPR technologies that are capable of detecting weak binding. It may also be used for high concentration biochemical screening.

Physical design

The library contains 352 compounds. The compounds are plated to facilitate screening with shape-diverse mixtures.

The library is distributed on four 96 well plates containing dry compound. Addition of 50µL of solvent will yield a 200mM solution. Compounds are soluble to 200mM in 100% DMSO.

We recommend that the plates are stored at room temperature. It is important that the area does not have a lot of moisture. For long term storage, a desiccator may be used.

Compound Properties

The mean compound properties in the Zenobia Therapeutics fragment library are below:

The compounds are all decorated rings - 'fragments of drugs' - similar to those described by Hartshorn et al (J.Med.Chem., 48, 403-413, 2005). Many of the compounds in the library have been seen as hits in experiments (biochemical and crystallographic screens). Useful Software

Crystallography The crystallographic model-building program MIFit includes the MIExpert system for pipelined structure solution calculations (using CCP4 software) that can be setup to calculate pre-refined electron density maps for a series of data set. The molecular viewer is convenient for displaying multiple models/maps in either the same canvas or in independent windows. This can be helpful for comparing related data sets.

This is an OpenSource software with precompiled binaries for Linux and Windows from: http://code.google.com/p/mifit/

Chemistry

The Hyleos ChemFileBrowser is a simple and convenient tool for scrolling through SD files containing compound structure data.

A free version of this software is available from: http://www.hyleos.net

Specific instructions for crystallographic screening

Crystallographic fragment screening is usually performed using groups of 3-10 compounds in a mixture in order to increase the efficiency of the experiment, albeit at a cost of reducing the effective concentration of each compound in the crystal. In order to maximize the chance of detecting binding, the highest possible effective concentration of compound in soaking (i.e. for crystal, mother liquor and compound mixture) should be used. The optimal number of compounds in a group depends on the tolerance of the crystal for DMSO (since this limits the amount of mixture solution with which the crystal may be soaked) and the target concentration of individual fragments in the crystal.

To help conduct the screening experiment, the compounds are organized into shape-diverse groups of eight compounds as the columns in each plate. The PDB file ‘Plate_plans’ contains the plate plans showing the compound’s Zenobia ID numbers as a function of well location. The chemical images (PNG files) are also sorted by mixtures into the folders within the Mixture_images folder. For example, the folder Mixture_images/plate1/1group1 contains images for the eight compounds in group 1 (column 1) of plate 1. Using the ‘View/thumbnails’ display mode is a convenient way for examining all of the molecules in the folder at the same time. A parallel set of folders contains PDB format files containing the three-dimensional structures of the compounds. These may be useful for docking into electron density difference maps in order to evaluate the identity of the compound from a mixture ‘hit’.

Since the compounds are pre-plated to become stock solutions with DMSO at 200mM concentration, mixing equi-volumes for a group of eight compounds leads to an effective concentration of 25mM of each compound in the mixture solution. Assuming a crystal tolerance for ~10% DMSO this leads to an effective concentration of ~2.5mM of each compound in the crystal. For DMSO-intolerant crystals or to increase the compound concentration we suggest ‘folding’ the columns and soaking crystals in two groups, containing four compounds each. Published work from several research groups indicates that many small fragment hits from crystallographic screening experiments bind with IC50 values in the range 0.1-1mM. However, it is sometimes observed that the compound density in the electron density>

map is clearer in follow-up single soak experiments than in the initial mixture screen. This suggests that fragment binding constants are in a regime where there is a trade-off between the desire to soak crystals at the higher concentrations possible in the smaller groups to detect the weaker binders and reducing data collection requirements by soaking in relatively large mixture sets.

Some hints for successful crystallographic fragment screening experiments

1. Before starting a fragment screening campaign it will be useful to assess the crystal’s tolerance for DMSO by soaking a crystal in various quantities of pure DMSO (without any dissolved compounds) and collecting sample diffraction data. Many crystals will tolerate ~10% DMSO by volume. Lower DMSO tolerance may mean that you should conduct fragment screening experiments in smaller mixtures (four compounds rather than eight) in order to maintain the concentration of each compound in the crystal. To increase the chance of observing a bound fragment the crystals should be soaked with as large a volume of DMSO as possible while maintaining reproducible diffraction across most crystals.

2. The length of time needed for effective soaking may be case dependent. Although small molecules should diffuse through a protein crystal extremely rapidly (probably within a few seconds) it might take significantly longer for specific binding in a target site to occur if this requires some kinetic event. Once a bound compound has been identified it is possible to evaluate this factor by soaking crystals for varying lengths of time. One report (Bosch et al, J.Med.Chem., 49, 5939-5946, 2006) indicates that binding may be rapid (<10s) and suggest that a ‘quick dip’ methodology is viable for crystals that do not survive prolonged soaking. For initial experiments, soaking for several minutes might be tried.

3. When starting a fragment screening campaign it is very useful to obtain a positive result in order to validate the soaking protocol and establish the possibility of success. If a small compound is known or expected to bind to the protein (for example, adenine will often bind to proteins that typically contain ADP) this may be used in preliminary experiments. Hit rates in many successful fragment screening projects are in the range 1-4% of total compounds screened so failure to detect any bound compounds after screening ~100 compounds may be a warning sign that either the soaking protocol is not working or the target site is not drugable.
4. DMSO or cryo-protectant molecules sometimes bind and block the target sites for fragment screening, with some variations between experiments. If several fragment screening experiments lead to suspiciously similar density features these molecules should be considered as accounting for the densities.

5. Occasionally a compound may crack the soaking crystals. If this occurs one approach is to divide the mixture group in half and repeat the soaking with two subsets. In this way the injurious compound will be partially isolated and identified and the other compounds in a shape-diverse mixture may still be recovered and included in the crystal fragment screen.

6. Small molecules may diffuse through crystals extremely rapidly (few seconds) and entry in to or out of the specific binding site might be quite rapid. To avoid risk of losing compounds from the crystal during freezing we suggest that the soaking mixture is included in the cryo-mixture at the same concentration as is used in soaking.

7. It may be useful to verify the identity of a hit from a mixture soaking experiment with a single-soak experiment. Since the single soak experiment may be performed with the compound at a higher effective concentration than the mixture experiment it may sometimes also provide a more interpretable density map with data at higher resolution and/or the compound at higher occupancy. If you intend to design elaborated compounds based on fragment hits (a relatively expensive undertaking) it is advisable to be certain of the fragment binding.