Beyond mere diversity: tailoring combinatorial libraries for drug discovery

Combinatorial library design attempts to choose the best set of substituents for a combinatorial synthetic scheme to maximize the chances of finding a useful compound, such as a drug lead. Initial efforts were focused primarily on maximizing diversity, perhaps allowing some bias by the inclusion of a small, fixed set of pharmacophoric substituents. However, many factors besides diversity impact good library design for drug discovery. A library can be better "tailored" by assigning the candidate substituents to categories such as polar, pharmacophoric, rigid, low molecular weight, and expensive. Stratified sampling by successive steps of D-optimal design generates diverse designs which are also consistent with desirable profiles of these properties. Comparing the diversity scores among design profiles reveals the tradeoffs between diversity, physical property distributions, synthetic difficulty, expense, and pharmacophoric bias. The diversity scores can be calibrated by scoring the best designs from subsets of the candidates made either from specific classes of substituents or by randomly eliminating candidates. This procedure shows how poor random designs are compared even to highly biased optimal designs. Library design requires a synergistic effort between computational and synthetic medicinal chemists, so specialized interactive software has been developed to integrate substructure searching, display, and statistical experimental design to facilitate this interaction for the effective design of well-tailored libraries.     

Representative subset selection

Fast development of analytical techniques enable to acquire huge amount of data. Large data sets are difficult to handle and therefore, there is a big interest in designing a subset of the original data set, which preserves the information of the original data set and facilitates the computations. There are many subset selection methods and their choice depends on the problem at hand. The two most popular groups of subset selection methods are uniform designs and cluster-based designs. Among the methods considered in this paper there are uniform designs, such as those proposed by Kennard and Stone, OptiSim, and cluster-based designs applying K-means technique and density based spatial clustering of applications with noise (DBSCAN). Additionally, a new concept of the subset selection with K-means is introduced.     

A fast exchange algorithm for designing focused libraries in lead optimization

Combinatorial chemistry is widely used in drug discovery. Once a lead compound has been identified, a series of R-groups and reagents can be selected and combined to generate new potential drugs. The combinatorial nature of this problem leads to chemical libraries containing usually a very large number of virtual compounds, far too large to permit their chemical synthesis. Therefore, one often wants to select a subset of "good" reagents for each R-group of reagents and synthesize all their possible combinations. In this research, one encounters some difficulties. First, the selection of reagents has to be done such that the compounds of the resulting sublibrary simultaneously optimize a series of chemical properties. For each compound, a desirability index, a concept proposed by Harrington,(20) is used to summarize those properties in one fitness value. Then a loss function is used as objective criteria to globally quantify the quality of a sublibrary. Second, there are a huge number of possible sublibraries, and the solutions space has to be explored as fast as possible. The WEALD algorithm proposed in this paper starts with a random solution and iterates by applying exchanges, a simple method proposed by Fedorov(13) and often used in the generation of optimal designs. Those exchanges are guided by a weighting of the reagents adapted recursively as the solutions space is explored. The algorithm is applied on a real database and reveals to converge rapidly. It is compared to results given by two other algorithms presented in the combinatorial chemistry literature: the Ultrafast algorithm of D. Agrafiotis and V. Lobanov and the Piccolo algorithm of W. Zheng et al.     

Virtual darwinian drug design: QSAR inverse problem, virtual combinatorial chemistry, and computational screening

The generation of diversity and its further selection by an external system is a common mechanism for the evolution of the living species and for the current drug design methods. This assumption allows us to label the methods based on generation and selection of molecular diversity as "Darwinian" ones, and to distinguish them from the structure-based, structure-modulation approaches. An example of a Darwinian method is the inverse QSAR. It consists of the computational generation of candidate chemical structures and their selection according to a previously established QSAR model. New trends in the field of combinatorial chemical syntheses comprise the concepts of virtual combinatorial synthesis and virtual or computational screening. Virtual combinatorial synthesis, closely related to inverse QSAR, can be defined as the computational simulation of the generation of new chemical structures by using a combinatorial strategy to generate a virtual library. Virtual screening is the selection of chemical structures having potential desirable properties from a database or virtual library in order to be synthesized and assayed. This review is mainly focused on graph theoretical drug design approaches, but a survey with key references is provided that covers other simulation methods.     

GLARE: a new approach for filtering large reagent lists in combinatorial library design using product properties

We present a novel computer algorithm, called GLARE (Global Library Assessment of REagents), that addresses the issue of optimal reagent selection in combinatorial library design. This program reduces or eliminates the time a medicinal chemist spends examining reagents which a priori cannot be part of a "good" library. Our approach takes the large reagent sets returned by standard chemical database queries and produces often considerably reduced reagent sets that are well-behaved with respect to a specific template. The pruning enforces "goodness" constraints such as the Lipinski rule of five on the product properties such that any reagent selection from the resulting sets produces only "good" products. The algorithm we implemented has three important features: (i) As opposed to genetic algorithms or other stochastic algorithms, GLARE uses a deterministic greedy procedure that smoothly filters out nonviable reagents. (ii) The pruning method can be biased to produce reagent sets with a balanced size, conserving proportionally more reagents in smaller sets. (iii) For very large combinatorial libraries, a partitioning scheme allows libraries as large as 10(12) to be evaluated in 0.25 s on an IBM AMD Opteron processor. This algorithm is validated on a diverse set of 12 libraries. The results that we obtained show an excellent compliance to the product property requirements and very fast timings.     

Boosted leave-many-out cross-validation: the effect of training and test set diversity on PLS statistics

It is becoming increasingly common in quantitative structure/activity relationship (QSAR) analyses to use external test sets to evaluate the likely stability and predictivity of the models obtained. In some cases, such as those involving variable selection, an internal test set – i.e., a cross-validation set – is also used. Care is sometimes taken to ensure that the subsets used exhibit response and/or property distributions similar to those of the data set as a whole, but more often the individual observations are simply assigned `at random.' In the special case of MLR without variable selection, it can be analytically demonstrated that this strategy is inferior to others. Most particularly, D-optimal design performs better if the form of the regression equation is known and the variables involved are well behaved. This report introduces an alternative, non-parametric approach termed `boosted leave-many-out' (boosted LMO) cross-validation. In this method, relatively small training sets are chosen by applying optimizable k-dissimilarity selection (OptiSim) using a small subsample size (k = 4, in this case), with the unselected observations being reserved as a test set for the corresponding reduced model. Predictive errors for the full model are then estimated by aggregating results over several such analyses. The countervailing effects of training and test set size, diversity, and representativeness on PLS model statistics are described for CoMFA analysis of a large data set of COX2 inhibitors.     

Combinatorial chemistry: libraries from libraries, the art of the diversity-oriented transformation of resin-bound peptides and chiral polyamides to low molecular weight acyclic and heterocyclic compounds

Combinatorial chemistry has deeply impacted the drug discovery process by accelerating the synthesis and screening of large numbers of compounds having therapeutic and/or diagnostic potential. These techniques offer unique enhancement in the potential identification of new and/or therapeutic candidates. Our efforts over the past 10 years in the design and diversity-oriented synthesis of low molecular weight acyclic and heterocyclic combinatorial libraries derived from amino acids, peptides, and/or peptidomimetics are described. Employing a "toolbox" of various chemical transformations, including alkylation, oxidation, reduction, acylation, and the use of a variety of multifunctional reagents, the "libraries from libraries" concept has enabled the continued development of an ever-expanding, structurally varied series of organic chemical libraries.     

Reagent Selector: using Synthon Analysis to visualize reagent properties and assist in combinatorial library design

Reagent Selector is an intranet-based tool that aids in the selection of reagents for use in combinatorial library construction. The user selects an appropriate reagent group as a query, for example, primary amines, and further refines it on the basis of various physicochemical properties, resulting in a list of potential reagents. The results of this selection process are, in turn, converted into synthons: the fragments or R-groups that are to be incorporated into the combinatorial library. The Synthon Analysis interface graphically depicts the chemical properties for each synthon as a function of the topological bond distance from the scaffold attachment point. Displayed in this fashion, the user is able to visualize the property space for the universe of synthons as well as that of the synthons selected. Ultimately, the reagent list that embodies the selected synthons is made available to the user for reagent procurement. Application of the approach to a sample reagent list for a G-protein coupled receptor targeted library is described.     

Immunoassays: biological tools for high throughput screening and characterisation of combinatorial libraries

In the demanding field of proteomics, there is an urgent need for affinity-catcher molecules to implement effective and high throughput methods for analysing the human proteome or parts of it. Antibodies have an essential role in this endeavour, and selection, isolation and characterisation of specific antibodies represent a key issue to meet success. Alternatively, it is expected that new, well-characterised affinity reagents generated in rapid and cost-effective manners will also be used to facilitate the deciphering of the function, location and interactions of the high number of encoded protein products. Combinatorial approaches combined with high throughput screening (HTS) technologies have become essential for the generation and identification of robust affinity reagents from biological combinatorial libraries and the lead discovery of active/mimic molecules in large chemical libraries. Phage and yeast display provide the means for engineering a multitude of antibody-like molecules against any desired antigen. The construction of peptide libraries is commonly used for the identification and characterisation of ligand-receptor specific interactions, and the search for novel ligands for protein purification. Further improvement of chemical and biological resistance of affinity ligands encouraged the "intelligent" design and synthesis of chemical libraries of low-molecular-weight bio-inspired mimic compounds. No matter what the ligand source, selection and characterisation of leads is a most relevant task. Immunological assays, in microtiter plates, biosensors or microarrays, are a biological tool of inestimable value for the iterative screening of combinatorial ligand libraries for tailored specificities, and improved affinities. Particularly, enzyme-linked immunosorbent assays are frequently the method of choice in a large number of screening strategies, for both biological and chemical libraries.     

Ultrafast algorithm for designing focused combinational arrays

A novel greedy algorithm for the design of focused combinatorial arrays is presented. The method is applicable when the objective function is decomposable to individual molecular contributions and makes use of a heuristic that allows the independent evaluation and ranking of candidate reagents in each variation site in the combinatorial library. The algorithm is extremely fast and convergent and produces solutions that are comparable to and often better than those derived from the substantially more elaborate and computationally intensive stochastic sampling techniques. Typical examples of design objectives that are amendable to this approach include maximum similarity to a known lead (or set of leads), maximum predicted activity according to some structure-activity or receptor binding model, containment within certain molecular property bounds, and many others.     

Assessing bias in total mercury results after removing a subsample from the bottle

U.S. EPA Method 1631 for total mercury (THg) analysis in water recommends that bromine monochloride (BrCl) be added to the original bottle in which the sample was collected, to draw into solution any Hg that may have adsorbed to the bottle walls. The method also allows for the removal of a subsample of water from the sample bottle for methylmercury (MeHg) analysis prior to adding BrCl. We have demonstrated that the removal of a subsample from the sample bottle prior to THg analysis can result in a positive concentration bias. The proposed mechanism for the bias is that ‘excess’ inorganic Hg, derived from the subsample that was removed from the bottle, adsorbs to the bottle walls and is then drawn into solution when BrCl is added. To test for this bias, we conducted an interlaboratory comparison study in which nine laboratories analysed water samples in fluorinated polyethylene (FLPE) bottles for THg after removing a subsample from the sample bottle, and analysed a replicate sample bottle from which no subsample was removed. We received seven complete data sets, or 63 unique sample pairs. The positive concentration bias between the bottles was significant when comparing all samples in aggregate (1.76 ± 0.53 ng/L after subsample removal, 1.57 ± 0.58 ng/L with no subsample removal, P < 0.05), however when comparing each of the three samples individually, the only significant bias was in the saline sample (Site UJ; 1.51 ± 0.31 ng/L after subsample removal, 1.32 ± 0.47 ng/L with no subsample removal, P < 0.05). Based on the findings presented here, we conclude that water chemistry, volume of water poured off, and the sample storage temperature explain some but not all of the observed bias, and we recommend collecting THg and MeHg samples in separate bottles whenever possible.     

Sensitivity analysis and other improvements to tailored combinatorial library design

"Tailoring" combinatorial libraries was developed several years ago as a very general and intuitive method to design diverse compound collections while controlling the profile of other pharmaceutically relevant properties. The candidate substituents were assigned to "categorical bins" according to their properties, and successive steps of D-optimal design were performed to generate diverse substituent sets consistent with required membership quotas from each bin. This serial algorithm was expedient to implement from existing D-optimal design codes, but was order-dependent and did not generally locate the very best possible design. A new "parallel" Fedorov search algorithm has now been implemented that can find the most diverse property-tailored design. An ambiguous mass penalty has been added, whereby most duplicate masses can be eliminated with little loss of library diversity. Sensitivity analysis has also been added to quantitatively explore the diversity trade-offs due to increasing or decreasing each specific kind of bias.     

A novel frequency distribution selection method for efficient plate layout of a diverse combinatorial library.

A deterministic method (frequency distribution method) for selecting compounds from a partitioned virtual combinatorial library for efficient synthesis is presented here. The method is based on reagent frequency analysis and can be applied to any library of molecules distributed in any given partitioned chemical space (cluster, cell-based, etc.). Compound selection by reagent frequency distribution can produce a unique, diverse set of molecules that adequately represents the library while requiring the least amount of compounds to be synthesized and minimizing the number of different reagents that must be used. This method also provides a practical solution to the configuration of plate layout. Because the method essentially identifies "expensive" regions in the chemical space to synthesize for a desired diversity or similarity coverage, decisions concerning the necessity to synthesize these compounds can be addressed. Minimum compound generation and efficient plate layout results in savings both in time of synthesis and cost of materials. This method always results in a discrete solution, which can be used for any given library size as well as any combination of reagents and is also readily adaptable to robotic automation.     

Characterisation of chlorinated hydrocarbons in a third generation candidate fish reference material (Limanda limanda) prepared under cryogenic conditions

Summary This paper describes the cryogenic preparation of 50 individual specimens of dab (Limanda limanda) from the Baltic Sea to produce a homogeneous fish material. 950 subsamples (8 g each) are stored in the gaseous phase over liquid nitrogen. Five series of analyses were randomly selected out of store. The results show that the obtained precision of 8 to 9% could be used as an indication unit for the deviation of the mean values of CHC concentrations from true values. If the analytical variability is 5%, the maximum inhomogeneity in the subsamples is 3–4%. It may be concluded that the subsamples of the candidate fish reference material (RM) are homogeneous enough for analytical purposes. Thus ESB (German Environmental Specimen Bank) material prepared in accordance with the standard operating procedure (SOP) [1] are valuable and appropriate reference materials. The prepared fish RM is suitable for the analysis of chlorinated hydrocarbons in fat containing marine samples.     

Accelerated screening of phage-display output with alkaline phosphatase fusions

When using multiple targets and libraries, selection of affinity reagents from phage-displayed libraries is a relatively time-consuming process. Herein, we describe an automation-amenable approach to accelerate the process by using alkaline phosphatase (AP) fusion proteins in place of the phage ELISA screening and subsequent confirmation steps with purified protein. After two or three rounds of affinity selection, the open reading frames that encode the affinity selected molecules (i.e., antibody fragments, engineered scaffold proteins, combinatorial peptides) are amplified from the phage or phagemid DNA molecules by PCR and cloned en masse by a Ligation Independent Cloning (LIC) method into a plasmid encoding a highly active variant of E. coli AP. This time-saving process identifies affinity reagents that work out of context of the phage and that can be used in various downstream enzyme linked binding assays. The utility of this approach was demonstrated by analyzing single-chain antibodies (scFvs), engineered fibronectin type III domains (FN3), and combinatorial peptides that were selected for binding to the Epsin N-terminal Homology (ENTH) domain of epsin 1, the c-Src SH3 domain, and the appendage domain of the gamma subunit of the clathrin adaptor complex, AP-1, respectively.     

Design of combinatorial libraries for the exploration of virtual hits from fragment space searches with LoFT

A case study is presented illustrating the design of a focused CDK2 library. The scaffold of the library was detected by a feature trees search in a fragment space based on reactions from combinatorial chemistry. For the design the software LoFT (Library optimizer using Feature Trees) was used. The special feature called FTMatch was applied to restrict the parts of the queries where the reagents are permitted to match. This way a 3D scoring function could be simulated. Results were compared with alternative designs by GOLD docking and ROCS 3D alignments.     

Polyaromatic Scavenger Reagents (PAHSR): A New Methodology for Rapid Purification in Solution-Phase Combinatorial Synthesis

A new method of purification of solution-phase combinatorial libraries has been developed. Development of a chemically inert polyaromatic anchor with a reactive "scavenger reagent" (PAHSR) allows unreacted reagents and impurities to be removed from a reaction by absorption of the PAHSR to charcoal and simple filtration.