Evaluation of similarity measures for searching the dictionary of natural products database

Similarity searches using combinations of seven different similarity coefficients and six different representations have been carried out on the Dictionary of Natural Products database. The objective was to discover if any special methods of searching apply to this database, which is very different in nature from the many synthetic databases that have been the subject of previous studies of similarity searching. Search effectiveness was assessed by a recall analysis of the search outputs from sets of pharmacologically active target structures. The different target sets produce exceptional but contradictory results for the Russell-Rao and Forbes coefficients, which have been shown to be due to a dependence on molecular size; these are the coefficients of choice in the case of large and small structures, respectively. Rankings from these results have been combined using a data fusion scheme and some small gains in performance were normally obtained by using substructural fingerprints and molecular holograms in combination with the Squared Euclidean or Tanimoto coefficients.     

CLIP: similarity searching of 3D databases using clique detection

This paper describes a program for 3D similarity searching, called CLIP (for Candidate Ligand Identification Program), that uses the Bron-Kerbosch clique detection algorithm to find those structures in a file that have large structures in common with a target structure. Structures are characterized by the geometric arrangement of pharmacophore points and the similarity between two structures calculated using modifications of the Simpson and Tanimoto association coefficients. This modification takes into account the fact that a distance tolerance is required to ensure that pairs of interatomic distances can be regarded as equivalent during the clique-construction stage of the matching algorithm. Experiments with HIV assay data demonstrate the effectiveness and the efficiency of this approach to virtual screening.     

Enhancing the effectiveness of virtual screening by fusing nearest neighbor lists: a comparison of similarity coefficients

This paper evaluates the effectiveness of various similarity coefficients for 2D similarity searching when multiple bioactive target structures are available. Similarity searches using several different activity classes within the MDL Drug Data Report and the Dictionary of Natural Products databases are performed using BCI 2D fingerprints. Using data fusion techniques to combine the resulting nearest neighbor lists we obtain group recall results which, in many cases, are a considerable improvement on standard average recall values obtained for individual structures. It is shown that the degree of improvement can be related to the structural diversity of the activity class that is searched for, the best results being found for the most diverse groups. The group recall of active compounds using subsets of the class is also investigated: for highly self-similar activity classes, the group recall improvement saturates well before the full activity class size is reached. A rough correlation is found between the relative improvement using the group recall and the square of the number of unique compounds available in all of the merged lists. The Tanimoto coefficient is found unambiguously to be the best coefficient to use for the recovery of active compounds using multiple targets. Furthermore, when using the Tanimoto coefficient, the "MAX" fusion rule is found to be more effective than the "SUM" rule for the combination of similarity searches from multiple targets. The use of group recall can lead to improved enrichment in database searches and virtual screening.     

Comparison of similarity coefficients for clustering and compound selection

Recent studies into the use of a selection of similarity coefficients, when applied to searches of chemical databases represented by binary fingerprints, have shown considerable variation in their retrieval performance and in the sets of compounds being retrieved. The main factor influencing performance is the density distribution of the bitstrings for the active class, a feature which is closely related to molecular size. If this is the case when these coefficients are applied to similarity searches, then we would expect considerable variation in performance when applied to dissimilarity methods, namely clustering and compound selection. Here we report on several studies which have been undertaken to investigate the relative performance of 13 association and correlation coefficients, which have been shown to exhibit complementary performance in similarity searches, when applied to hierarchical and nonhierarchical clustering methods and to a compound selection methodology. Results suggest that the correlation coefficients perform consistently well for clustering and compound selection, as does the Baroni-Urbani/Buser association coefficient. Surprisingly, these often outperform the Tanimoto coefficient, while the Simple Match (effectively the complement of the Squared Euclidean Distance) performs very poorly.     

Similarity by compression

We present a simple and effective method for similarity searching in virtual high-throughput screening, requiring only a string-based representation of the molecules (e.g., SMILES) and standard compression software, available on all modern desktop computers. This method utilizes the normalized compression distance, an approximation of the normalized information distance, based on the concept of Kolmogorov complexity. On representative data sets, we demonstrate that compression-based similarity searching can outperform standard similarity searching protocols, exemplified by the Tanimoto coefficient combined with a binary fingerprint representation and data fusion. Software to carry out compression-based similarity is available from our Web site at http://comp.chem.nottingham.ac.uk/download/zippity.     

Spatial chemical distance based on atomic property fields

Similarity of compound chemical structures often leads to close pharmacological profiles, including binding to the same protein targets. The opposite, however, is not always true, as distinct chemical scaffolds can exhibit similar pharmacology as well. Therefore, relying on chemical similarity to known binders in search for novel chemicals targeting the same protein artificially narrows down the results and makes lead hopping impossible. In this study we attempt to design a compound similarity/distance measure that better captures structural aspects of their pharmacology and molecular interactions. The measure is based on our recently published method for compound spatial alignment with atomic property fields as a generalized 3D pharmacophoric potential. We optimized contributions of different atomic properties for better discrimination of compound pairs with the same pharmacology from those with different pharmacology using Partial Least Squares regression. Our proposed similarity measure was then tested for its ability to discriminate pharmacologically similar pairs from decoys on a large diverse dataset of 115 protein–ligand complexes. Compared to 2D Tanimoto and Shape Tanimoto approaches, our new approach led to improvement in the area under the receiver operating characteristic curve values in 66 and 58% of domains respectively. The improvement was particularly high for the previously problematic cases (weak performance of the 2D Tanimoto and Shape Tanimoto measures) with original AUC values below 0.8. In fact for these cases we obtained improvement in 86% of domains compare to 2D Tanimoto measure and 85% compare to Shape Tanimoto measure. The proposed spatial chemical distance measure can be used in virtual ligand screening.     

Comparison of 2D similarity and 3D superposition. Application to searching a conformational drug database

In a database of about 2000 approved drugs, represented by 10(5) structural conformers, we have performed 2D comparisons (Tanimoto coefficients) and 3D superpositions. For one class of drugs the correlation between structural resemblance and similar action was analyzed in detail. In general Tanimoto coefficients and 3D scores give similar results, but we find that 2D similarity measures neglect important structural/funtional features. Examples for both over- and underestimation of similarity by 2D metrics are discussed. The required additional effort for 3D superpositions is assessed by implementation of a fast algorithm with a processing time below 0.01 s and a more sophisticated approach (0.5 s per superposition). According to the improvement of similarity detection compared to 2D screening and the pleasant rapidity on a desktop PC, full-atom 3D superposition will be an upcoming method of choice for library prioritization or similarity screening approaches.     

Fragment‐based similarity searching with infinite color space

Fragment‐based searching and abstract representation of molecular features through reduced graphs have separately been used for virtual screening. Here, we combine these two approaches and apply the algorithm RedFrag to virtual screens retrospectively and prospectively. It uses a new type of reduced graph that does not suffer from information loss during its construction and bypasses the necessity of feature definitions. Built upon chemical epitopes resulting from molecule fragmentation, the reduced graph embodies physico‐chemical and 2D‐structural properties of a molecule. Reduced graphs are compared with a continuous‐similarity‐distance‐driven maximal common subgraph algorithm, which calculates similarity at the fragmental and topological levels. The performance of the algorithm is evaluated by retrieval experiments utilizing precompiled validation sets. By predicting and experimentally testing ligands for endothiapepsin, a challenging model protease, the method is assessed in a prospective setting. Here, we identified five novel ligands with affinities as low as 2.08 μM. © 2015 Wiley Periodicals, Inc.     

Analysis and display of the size dependence of chemical similarity coefficients

We discuss the size-bias inherent in several chemical similarity coefficients when used for the similarity searching or diversity selection of compound collections. Limits to the upper bounds of 14 standard similarity coefficients are investigated, and the results are used to identify some exceptional characteristics of a few of the coefficients. An additional numerical contribution to the known size bias in the Tanimoto coefficient is identified. Graphical plots with respect to relative bit density are introduced to further assess the coefficients. Our methods reveal the asymmetries inherent in most similarity coefficients that lead to bias in selection, most notably with the Forbes and Russell-Rao coefficients. Conversely, when applied to the recently introduced Modified Tanimoto coefficient our methods provide support for the view that it is less biased toward molecular size than most. In this work we focus our discussion on fragment-based bit strings, but we demonstrate how our approach can be generalized to continuous representations.