Structure Ranking

A principal assumption of CSP by lattice energy minimization is that the observed crystal structure (or structures, where polymorphism occurs) corresponds to the lowest energy possible crystal packings of the molecule in question. Thus, the quahty of the model used to evaluate the energies of predicted crystal structures is key to the 

The most commonly applied method is the atom-atom potential, or force field, method for calculating lattice energies. Initially considering intermolecular interactions only, the total lattice energy is assessed as a sum over intermolecular interactions, which in turn are computed as a sum over atom-atom interactions, U  

The functional form of atom-atom interactions must contain terms describing the exchange-repulsion, the attractive dispersion interaction between atoms, and the electrostatic interaction  

Electrostatic interactions can be most simply modeled as the Coulomb interaction between partial atomic charges, while the repulsion-dispersion part is usually described by a Lennard-Jones or, more accurately, an exp-6 form, each of which contains parameters that must be fixed. High-quality empirically fitted parameter sets have been developed, where the atom-atom interactions are parameterized to reproduce the structures, sublimation enthalpies and, sometimes, further observable properties of organic molecular crystals [73,74]. Their use has been very effective in CSP. Nonempirical approaches to fitting intermolecular force fields, where the parameters are derived from quantum mechanical calculations, have occasionally been applied for CSP [75-78], but these are currently limited to small molecules, so currently lack relevance for typical pharmaceutical molecules. 

The main alternative approach to force field methods that has emerged in the past few years is the application of solid-state (periodic) DFT calculations for the lattice energy minimization and energetic evaluation of predicted crystal structures. The main weakness of DFT is its failure to account for the attractive dispersion interactions between 

---

A spreadsheet is opened with all structures ranked in a column. Select all rows with conformations and add energy scores ... 

Given two or more carboxylic acids with closely related structures, rank them in order of increasing (or decreasing) acidities (pKas). 

Figure 3 The frequency of predicting the crystal structure of the SB203386-HIV-1 protease complex in docking simulations with the ensemble of 6 protein conformations (a) and the ensemble of 32 protein conformations (b). The RMSD of the docked inhibitor conformations from the crystal structure ranked by energy in simulations with the ensemble of 6 protein conformations (c) and the ensemble of 32 protein conformations (d). The piecewise linear energy function is employed.

In trials with FTDOCK, it was found that the electrostatic term worked best as a binary filter. Complexes with unfavourable electrostatics are discarded and the remaining docked structures ranked by shape complementarity. For a given orientation of the movable molecule, the shape complementarity correlation function c is examined and the three most favourable solutions stored. After all orientations are sampled, the top set (typically 10000) of complexes are kept for subsequent examination. 

In judging the success of ranking methods, we should discard those entries that did not produce the structure in the first step. Instead, we ask how many of those that did generate the correct structure ranked it within their top three Category I success rates were 3 of 10 = 30% (IV), 5 of 11 = 45% (Vni), 0 of 4 = 0% (XI) and 4 of 10 = 40%... 

A particularly clear-cut result is obtained if structures are restricted to those of the correct molecular formula Cg Hj2 O2. This leaves us with four structure candidates, with the correct structure ranked top, with a match value of 0.60530 ... 

Draw all possible staggered and eclipsed conformations of l-bromo-2-chloroethane (see the followingball-and-stick model), using Newman projections. Underneath each structure, draw the corresponding dash-wedge and sawhorse structures. Rank the conformations in order of decreasing stability. 

Some of the top structures however, had extra side chains, or had side chains which were longer or shorter than those in the query. Searches based on the canonical root atom did not produce good results. Although the three existing occurrences of the query structure were ranked in the top five structures, the other structures ranked similar to the query were in fact much more complex, with much larger patterns and numbers of atoms. 

Nonbinary structural representations, such as sets of topological indices, are normally compared by means of the Euclidean distance, and the database structures ranked in order of increasing distance from, rather than decreasing similarity to, the target structure. An alternative approach to the calculation of molecular similarity when nonbinary strings are to be compared is discussed in Section 3.2,... 

Comprehensive computer-based systems of structure elucidation include, or are under development to include, capabilities in spectrum interpretation,. structure generation, and spectrum prediction and comparison. Considerable variation in approach by different investigators can be found, but the goal is the same, to produce, at best, a single structure of correct assignment and, at worst, a small, but exhaustive set of alternative structures ranked in order of decreasing probability of being correct. 

---