Similarity searching in files of three-dimensional chemical structures: Representation and searching of molecular electrostatic potentials using field-graphs

This paper reports a method for the identification of those molecules in a database of rigid 3D structures with molecular electrostatic potential (MEP) grids that are most similar to that of a user-defined target molecule. The most important features of an MEP grid are encoded in field-graphs, and a target molecule is matched against a database molecule by a comparison of the corresponding field-graphs. The matching is effected using a maximal common subgraph isomorphism algorithm, which provides an alignment of the target molecule's field- graph with those of each of the database molecules in turn. These alignments are used in the second stage of the search algorithm to calculate the intermolecular MEP similarities. Several different ways of generating field-graphs are evaluated, in terms of the effectiveness of the resulting similarity measures and of the associated computational costs. The most appropriate procedure has been implemented in an operational system that searches a corporate database, containing ca. 173,000 3D structures.     

Calculation of hydrophobic parameters directly from three-dimensional structures using comparative molecular field analysis

Summary Capacity ratio (log k') values, which are a measure of hydrophobicity, were calculated directly from the three-dimensional structures of 17 furans and 54 triazines using the comparative molecular field analysis approach. The H₂O probe and the GRID force field, including hydrogen-bond potentials, yielded excellent correlations with the log k' values. Moreover, the predicted values of log k' from 14 additional triazine analogs showed excellent agreement with log k' values reported in the literature. Similar results were obtained for the octanol-water partition coefficients (log P) of 17 furans.     

Clustering files of chemical structures using the fuzzy k-means clustering method

This paper evaluates the use of the fuzzy k-means clustering method for the clustering of files of 2D chemical structures. Simulated property prediction experiments with the Starlist file of logP values demonstrate that use of the fuzzy k-means method can, in some cases, yield results that are superior to those obtained with the conventional k-means method and with Ward's clustering method. Clustering of several small sets of agrochemical compounds demonstrate the ability of the fuzzy k-means method to highlight multicluster membership and to identify outlier compounds, although the former can be difficult to interpret in some cases.     

Similarity searching in databases of three-dimensional molecules and macromolecules.

This paper discusses algorithmic techniques for measuring the degree of similarity between pairs of three-dimensional (3-D) chemical molecules represented by interatomic distance matrices. A comparison of four methods for the calculation of 3-D structural similarity suggests that the most effective one is a procedure that identifies pairs of atoms, one from each of the molecules that are being compared, that lie at the center of geometrically-related volumes of 3-D space. This atom mapping method enables the calculation of a wide range of types of intermolecular similarity coefficient, including measures that are based on physicochemical data. Massively-parallel implementations of the method are discussed, using the AMT Distributed Array Processor, that achieve a substantial increase in performance when compared with a sequential implementation on a UNIX workstation. Current work involves the use of angular information and the extension of the method to field-based similarity searching. Similarity searching in 3-D macromolecules is effected by the use of a maximal common subgraph (MCS) isomorphism algorithm with a novel, graph-based representation of the tertiary structures of proteins. This algorithm is being used to identify similarities between the 3-D structures of proteins in the Brookhaven Protein Data Bank; its use is exemplified by searches involving the NAD-binding fold motif.     

Prediction of pesticide acute toxicity using two-dimensional chemical descriptors and target species classification

Previous modelling of the median lethal dose (oral rat LD₅₀) has indicated that local class-based models yield better correlations than global models. We evaluated the hypothesis that dividing the dataset by pesticidal mechanisms would improve prediction accuracy. A linear discriminant analysis (LDA) based-approach was utilized to assign indicators such as the pesticide target species, mode of action, or target species - mode of action combination. LDA models were able to predict these indicators with about 87% accuracy. Toxicity is predicted utilizing the QSAR model fit to chemicals with that indicator. Toxicity was also predicted using a global hierarchical clustering (HC) approach which divides data set into clusters based on molecular similarity. At a comparable prediction coverage (~94%), the global HC method yielded slightly higher prediction accuracy (r² = 0.50) than the LDA method (r² ~ 0.47). A single model fit to the entire training set yielded the poorest results (r² = 0.38), indicating that there is an advantage to clustering the dataset to predict acute toxicity. Finally, this study shows that whilst dividing the training set into subsets (i.e. clusters) improves prediction accuracy, it may not matter which method (expert based or purely machine learning) is used to divide the dataset into subsets.     

Similarity searching in databases of flexible 3D structures using smoothed bounded distance matrices

This paper describes a method for calculating the similarity between pairs of chemical structures represented by 3D molecular graphs. The method is based on a graph matching procedure that accommodates conformational flexibility by using distance ranges between pairs of atoms, rather than fixing the atom pair distances. These distance ranges are generated using triangle and tetrangle bound smoothing techniques from distance geometry. The effectiveness of the proposed method in retrieving other compounds of like biological activity is evaluated, and the results are compared with those obtained from other, 2D-based methods for similarity searching.     

Nondestructive three-dimensional analysis of layered polymer structures with chemical imaging

Three-dimensional (3D) chemical information was obtained by means of a combination of two-dimensional attenuated total reflection Fourier transform infrared (ATR-FT-IR) imaging with a focal plane array detector and variable angle depth profiling. Since the penetration depth of the evanescent wave in ATR spectroscopy is not limited by diffraction, it was possible to resolve thin sandwiched polymer layers nondestructively within a stack of polymer layers. Chemical images were obtained from layers of different thickness of the laminate by moving a custom-made aperture to specific positions on the condenser lens of the ATR accessory. Sequences of absorption images detect the successive appearance of thin, buried layers of polybutylmethacrylate (d(PBMA) = 400 nm) and polycarbonate (d(TMPC) = 300 nm) in different depths of the stack of polymer layers. The depth resolution of variable-angle ATR-FT-IR imaging is sufficiently high to detect surface roughness at the interface between different polymer layers. Two different stacks of polymers with reordered sandwich-layers were imaged simultaneously, demonstrating the potential of variable angle ATR-FT-IR for 3D-imaging of a sample with xyz-heterogeneity, which can be a powerful analytical technique for materials science and biomedical research.     

ALADDIN: An integrated tool for computer-assisted molecular design and pharmacophore recognition from geometric,steric, and substructure searching of three-dimensional molecular structures

Summary ALADDIN is a computer program for the design or recognition of compounds that meet geometric, steric, and substructural criteria. ALADDIN searches a database of three-dimensional structures, marks atoms that meet substructural criteria, evaluates geometric criteria, and prepares a number of files that are input for molecular modification and coordinate generation as well as for molecular graphics. Properties calculated from the three-dimensional structure are described by either properties calculated from the molecule itself or from the molecule as compared to a reference molecule and associated surfaces. ALADDIN was used to design analogues to probe a bioactive conformation of a small molecule and a peptide, to test alternative superposition rules for receptor mapping of the D2 dopamine receptor, to recognize unexpected D2 dopamine agonist activity of existing compounds, and to design compounds to fit a binding site on a protein of known structure. We have found that series designed by ALADDIN show much more subtle variation in shape than do those designed by traditional methods and that compounds can be designed to be very close matches to the objective.