The concept of template-based de novo design from drug-derived molecular fragments and its application to TAR RNA.

Principles of fragment-based molecular design are presented and discussed in the context of de novo drug design. The underlying idea is to dissect known drug molecules in fragments by straightforward pseudo-retro-synthesis. The resulting building blocks are then used for automated assembly of new molecules. A particular question has been whether this approach is actually able to perform scaffold-hopping. A prospective case study illustrates the usefulness of fragment-based de novo design for finding new scaffolds. We were able to identify a novel ligand disrupting the interaction between the Tat peptide and TAR RNA, which is part of the human immunodeficiency virus (HIV-1) mRNA. Using a single template structure (acetylpromazine) as reference molecule and a topological pharmacophore descriptor (CATS), new chemotypes were automatically generated by our de novo design software Flux. Flux features an evolutionary algorithm for fragment-based compound assembly and optimization. Pharmacophore superimposition and docking into the target RNA suggest perfect matching between the template molecule and the designed compound. Chemical synthesis was straightforward, and bioactivity of the designed molecule was confirmed in a FRET assay. This study demonstrates the practicability of de novo design to generating RNA ligands containing novel molecular scaffolds.

Molecular query language (MQL)--a context-free grammar for substructure matching.

We have developed a Java library for substructure matching that features easy-to-read syntax and extensibility. This molecular query language (MQL) is grounded on a context-free grammar, which allows for straightforward modification and extension. The formal description of MQL is provided in this paper. Molecule primitives are atoms, bonds, properties, branching, and rings. User-defined features can be added via a Java interface. In MQL, molecules are represented as graphs. Substructure matching was implemented using the Ullmann algorithm because of favorable run-time performance. The Ullmann algorithm carries out a fast subgraph isomorphism search by combining backtracking with effective forward checking. MQL software design was driven by the aim to facilitate the use of various cheminformatics toolkits. Two Java interfaces provide a bridge from our MQL package to an external toolkit: the first one provides the matching rules for every feature of a particular toolkit; the second one converts the found match from the internal format of MQL to the format of the external toolkit. We already implemented these interfaces for the Chemistry Development Toolkit.

Flux (2): comparison of molecular mutation and crossover operators for ligand-based de novo design.

We implemented a fragment-based de novo design algorithm for a population-based optimization of molecular structures. The concept is grounded on an evolution strategy with mutation and crossover operators for structure breeding. Molecular building blocks were obtained from the pseudo-retrosynthesis of a collection of pharmacologically active compounds following the RECAP principle. The influence of mutation and crossover on the course of optimization was assessed in redesign studies using known drugs as template structures. A topological atom-pair descriptor grounded on potential pharmacophore points was used as a molecular descriptor, and the Manhattan distance between the template and candidate molecules served as a fitness function. Exclusive use of the crossover operator yielded few unique compounds and often resulted in premature convergence of the optimization process, whereas exclusive use of the mutation operator resulted in diverse high-quality structures. Combinations of crossover and mutation yielded the overall best results. The majority of the designed structures exhibit a chemically reasonable architecture; chiral centers are rare, and unfavorable connections of building blocks are infrequent. We conclude that this fragment-based design principle is suited as an idea generator for the automated design of novel leadlike molecules.

A pseudo-ligand approach to virtual screening.

A virtual screening method is presented that is grounded on a receptor-derived pharmacophore model termed "virtual ligand" or "pseudo-ligand". The model represents an idealized constellation of potential ligand sites that interact with residues of the binding pocket. For rapid virtual screening of compound libraries the potential pharmacophore points of the virtual ligand are encoded as an alignment-free correlation vector, avoiding spatial alignment of pharmacophore features between the pharmacophore query (i.e., the virtual ligand) and the candidate molecule. The method was successfully applied to retrieving factor Xa inhibitors from a Ugi three-component combinatorial library, and yielded high enrichment of actives in a retrospective search for cyclooxygenase-2 (COX-2) inhibitors. The approach provides a concept for "de-orphanizing" potential drug targets and identifying ligands for hitherto unexplored or allosteric binding pockets.

Flux (1): a virtual synthesis scheme for fragment-based de novo design.

It is demonstrated that the fragmentation of druglike molecules by applying simplistic pseudo-retrosynthesis results in a stock of chemically meaningful building blocks for de novo molecule generation. A stochastic search algorithm in conjunction with ligand-based similarity scoring (Flux: fragment-based ligand builder reaxions) facilitated the generation of new molecules using a single known reference compound as a template. This molecule assembly method is applicable in the absence of receptor-structure information. In a case study, we used imantinib (Gleevec) and a Factor Xa inhibitor as the reference structures. The algorithm succeeded in redesigning the templates from scratch and suggested several alternative molecular structures. The resulting designed molecules were chemically reasonable and contained essential substructure motifs. A comparison of molecular descriptors suggests that holographic descriptors might be advantageous over binary fingerprints for ligand-based de novo design.

Computer-based de novo design of drug-like molecules.

Ever since the first automated de novo design techniques were conceived only 15 years ago, the computer-based design of hit and lead structure candidates has emerged as a complementary approach to high-throughput screening. Although many challenges remain, de novo design supports drug discovery projects by generating novel pharmaceutically active agents with desired properties in a cost- and time-efficient manner. In this review, we outline the various design concepts and highlight current developments in computer-based de novo design.

Advances in the prediction of protein targeting signals.

Enlarged sets of reference data and special machine learning approaches have improved the accuracy of the prediction of protein subcellular localization. Recent approaches report over 95% correct predictions with low fractions of false-positives for secretory proteins. A clear trend is to develop specifically tailored organism- and organelle-specific prediction tools rather than using one general method. Focus of the review is on machine learning systems, highlighting four concepts: the artificial neural feed-forward network, the self-organizing map (SOM), the Hidden-Markov-Model (HMM), and the support vector machine (SVM).

Comparison of support vector machine and artificial neural network systems for drug/nondrug classification.

Support vector machine (SVM) and artificial neural network (ANN) systems were applied to a drug/nondrug classification problem as an example of binary decision problems in early-phase virtual compound filtering and screening. The results indicate that solutions obtained by SVM training seem to be more robust with a smaller standard error compared to ANN training. Generally, the SVM classifier yielded slightly higher prediction accuracy than ANN, irrespective of the type of descriptors used for molecule encoding, the size of the training data sets, and the algorithm employed for neural network training. The performance was compared using various different descriptor sets and descriptor combinations based on the 120 standard Ghose-Crippen fragment descriptors, a wide range of 180 different properties and physicochemical descriptors from the Molecular Operating Environment (MOE) package, and 225 topological pharmacophore (CATS) descriptors. For the complete set of 525 descriptors cross-validated classification by SVM yielded 82% correct predictions (Matthews cc = 0.63), whereas ANN reached 80% correct predictions (Matthews cc = 0.58). Although SVM outperformed the ANN classifiers with regard to overall prediction accuracy, both methods were shown to complement each other, as the sets of true positives, false positives (overprediction), true negatives, and false negatives (underprediction) produced by the two classifiers were not identical. The theory of SVM and ANN training is briefly reviewed.

Comparison of correlation vector methods for ligand-based similarity searching.

Correlation vector methods were tested for their usefulness in ligand-based virtual screening. Three molecular descriptors--two based on potential pharmacophore points and one on partial atom charges--and three similarity measures--the Manhattan distance, the Euclidian distance and the Tanimoto coefficient--were compared. The alignment-free descriptors seem to be particularly applicable when a course-grain filtering of data sets is required in combination with a high execution speed. Significant enrichment of actives was obtained by retrospective analysis. The cumulative percentages for all three descriptors allow for the retrieval of up to 78% of the active molecules in the first five percent of the reference database. Different descriptors retrieved only weakly overlapping sets of active molecules among the top-ranking compounds. If a single similarity index is to be used, the Manhattan distance seems to be particularly applicable. Generally, none of the three different descriptors tested in this study clearly outperformed the others. The suitability of a descriptor critically depends on the ligand-receptor interaction under investigation. For ligand-based similarity searching it is recommended to exploit several descriptors in parallel.