Understanding Data Noise and Uncertainty through Analysis of Replicate Samples in DNA-Encoded Library Selection.

By analyzing data sets of replicate DNA-Encoded Library (DEL) selections, an approach for estimating the noise level of the experiment has been developed. Using a logarithm transformation of the number of counts associated with each compound and a subset of compounds with the highest number of counts, it is possible to assess the quality of the data through normalizing the replicates and use this same data to estimate the noise in the experiment. The noise level is seen to be dependent on sequencing depth as well as specific selection conditions. The noise estimation is independent of any cutoff used to remove low frequency compounds from the data analysis. The removal of compounds with only 1-5 read counts greatly reduces some of the challenges encountered in DEL data analysis as it can reduce the data set by greater than 100-fold without impacting the interpretation of the results.

Designing DNA Encoded Libraries of Diverse Products in a Focused Property Space.

DNA encoded libraries (DEL) are being used as a complement or alternative to traditional high throughput screening (HTS). To maximize the chances of finding chemically attractive lead material that is appropriate for medicinal chemistry optimization, for example, in the Rule of Five compliant chemistry space, it is important to design DEL library compounds such that they are highly diverse and fall within a desired property space. Currently available library design methods can be classified as either monomer-based or product-based. As monomers may undergo significant structural changes when participating in a reaction, monomer based design can provide a poor representation of the properties of resultant DEL products. However, product-based design introduces a technical obstacle due to the enormous chemical design space for many DELs. Here a new method for monomer based selections is described using representative sublibraries as surrogates for fully enumerated DEL property-based optimization. Through a series of rational and systematic library enumerations and property calculations, building-block representatives are identified and representative sublibraries are defined to drive the optimization process. A published data set for a triazine library was used to demonstrate the effectiveness of the multiple objective optimization for six properties. All of the evaluated properties for the designed library are shown to consistently shift toward the desired property distribution as driven by the design criteria.

Novel Methods for Prioritizing "Close-In" Analogs from Structure-Activity Relationship Matrices.

Here we describe the development of novel methods for compound evaluation and prioritization based on the structure-activity relationship matrix (SARM) framework. The SARM data structure allows automatic and exhaustive extraction of SAR patterns from data sets and their organization into a chemically intuitive scaffold/functional-group format. While SARMs have been used in the retrospective analysis of SAR discontinuity and identifying underexplored regions of chemistry space, there have been only a few attempts to apply SARMs prospectively in the prioritization of "close-in" analogs. In this work, three new ways of prioritizing virtual compounds based on SARMs are described: (1) matrix pattern-based prioritization, (2) similarity weighted, matrix pattern-based prioritization, and (3) analysis of variance based prioritization (ANV). All of these methods yielded high predictive power for six benchmark data sets (prediction accuracy R2 range from 0.63 to 0.82), yielding confidence in their application to new design ideas. In particular, the ANV method outperformed the previously reported SARM based method for five out of the six data sets tested. The impact of various SARM parameters were investigated and the reasons why SARM-based compound prioritization methods provide higher predictive power are discussed.

Improved nucleic acid descriptors for siRNA efficacy prediction.

Although considerable progress has been made recently in understanding how gene silencing is mediated by the RNAi pathway, the rational design of effective sequences is still a challenging task. In this article, we demonstrate that including three-dimensional descriptors improved the discrimination between active and inactive small interfering RNAs (siRNAs) in a statistical model. Five descriptor types were used: (i) nucleotide position along the siRNA sequence, (ii) nucleotide composition in terms of presence/absence of specific combinations of di- and trinucleotides, (iii) nucleotide interactions by means of a modified auto- and cross-covariance function, (iv) nucleotide thermodynamic stability derived by the nearest neighbor model representation and (v) nucleic acid structure flexibility. The duplex flexibility descriptors are derived from extended molecular dynamics simulations, which are able to describe the sequence-dependent elastic properties of RNA duplexes, even for non-standard oligonucleotides. The matrix of descriptors was analysed using three statistical packages in R (partial least squares, random forest, and support vector machine), and the most predictive model was implemented in a modeling tool we have made publicly available through SourceForge. Our implementation of new RNA descriptors coupled with appropriate statistical algorithms resulted in improved model performance for the selection of siRNA candidates when compared with publicly available siRNA prediction tools and previously published test sets. Additional validation studies based on in-house RNA interference projects confirmed the robustness of the scoring procedure in prospective studies.

HELM: a hierarchical notation language for complex biomolecule structure representation.

When biological macromolecules are used as therapeutic agents, it is often necessary to introduce non-natural chemical modifications to improve their pharmaceutical properties. The final products are complex structures where entities such as proteins, peptides, oligonucleotides, and small molecule drugs may be covalently linked to each other, or may include chemically modified biological moieties. An accurate in silico representation of these complex structures is essential, as it forms the basis for their electronic registration, storage, analysis, and visualization. The size of these molecules (henceforth referred to as "biomolecules") often makes them too unwieldy and impractical to represent at the atomic level, while the presence of non-natural chemical modifications makes it impossible to represent them by sequence alone. Here we describe the Hierarchical Editing Language for Macromolecules ("HELM") and demonstrate its utility in the representation of structures such as antisense oligonucleotides, short interference RNAs, peptides, proteins, and antibody drug conjugates.

Molecular dynamics simulation and binding energy calculation for estimation of oligonucleotide duplex thermostability in RNA-based therapeutics.

For oligonucleotide-based therapeutics, a thorough understanding of the thermodynamic properties of duplex formation is critical to developing stable and potent drugs. For unmodified small interfering RNA (siRNA), DNA antisense oligonucleotide (AON) and locked nucleic acid (LNA), DNA/LNA modified oligonucleotides, nearest neighbor (NN) methods can be effectively used to quickly and accurately predict duplex thermodynamic properties such as melting point. Unfortunately, for chemically modified olignonucleotides, there has been no accurate prediction method available. Here we describe the potential of estimating melting temperature (T(m)) for nonstandard oligonucleotides by using the correlation of the experimental T(m) with the calculated duplex binding energy (BE) for oligonucleotides of a given length. This method has been automated into a standardized molecular dynamics (MD) protocol through Pipeline Pilot (PP) using the CHARMm component in Discovery Studio (DS). Results will be presented showing the correlation of the predicted data with experiment for both standard and chemically modified siRNA and AON.

PFRED: A computational platform for siRNA and antisense oligonucleotides design.

PFRED a software application for the design, analysis, and visualization of antisense oligonucleotides and siRNA is described. The software provides an intuitive user-interface for scientists to design a library of siRNA or antisense oligonucleotides that target a specific gene of interest. Moreover, the tool facilitates the incorporation of various design criteria that have been shown to be important for stability and potency. PFRED has been made available as an open-source project so the code can be easily modified to address the future needs of the oligonucleotide research community. A compiled version is available for downloading at https://github.com/pfred/pfred-gui/releases/tag/v1.0 as a java Jar file. The source code and the links for downloading the precompiled version can be found at https://github.com/pfred.

High-throughput virtual screening of proteins using GRID molecular interaction fields.

A new computational algorithm for protein binding sites characterization and comparison has been developed, which uses a common reference framework of the projected ligand-space four-point pharmacophore fingerprints, includes cavity shape, and can be used with diverse proteins as no structural alignment is required. Protein binding sites are first described using GRID molecular interaction fields (GRID-MIFs), and the FLAP (fingerprints for ligands and proteins) method is then used to encode and compare this information. The discriminating power of the algorithm and its applicability for large-scale protein analysis was validated by analyzing various scenarios: clustering of kinase protein families in a relevant manner, predicting ligand activity across related targets, and protein-protein virtual screening. In all cases the results showed the effectiveness of the GRID-FLAP method and its potential use in applications such as identifying selectivity targets and tools/hits for new targets via the identification of other proteins with pharmacophorically similar binding sites.

The Identification of Pivotal Transcriptional Factors Mediating Cell Responses to Drugs With Drug-Induced Liver Injury Liabilities.

Drug-induced liver injury (DILI) is a leading cause of drug attrition during drug development and a common reason for drug withdrawal from the market. The poor predictability of conventional animal-based approaches necessitates the development of alternative testing approaches. A body of evidence associates DILI with the induction of stress-response genes in liver cells. Here, we set out to identify signal transduction pathways predominantly involved in the regulation of gene transcription by DILI drugs. To this end, we employed ATTAGENE's cell-based multiplexed reporter assay, the FACTORIAL transcription factor (TF), that enables quantitative assessment of the activity of multiple stress-responsive TFs in a single well of cells. Homogeneous reporter system enables quantitative functional assessment of multiple transcription factors. Nat. Methods 5, 253-260). Using this assay, we assessed TF responses of the human hepatoma cell line HepG2 to a panel of 64 drug candidates, including 23 preclinical DILI and 11 clinical DILI compounds and 30 nonhepatotoxic compounds from a diverse physicochemical property space. We have identified 16 TF families that specifically responded to DILI drugs, including nuclear factor (erythroid-derived 2)-like 2 antioxidant response element, octamer, hypoxia inducible factor 1 alpha, farnesoid-X receptor, TCF/beta-catenin, aryl hydrocarbon receptor, activator protein-1, E2F, early growth response-1, metal-response transcription factor 1, sterol regulatory element-binding protein, paired box protein, peroxisome proliferator-activated receptor, liver X receptor, interferone regulating factor, and P53, and 2 promoters that responded to multiple TFs (cytomegalovirus and direct repeat 3/vitamin D receptor). Some of TFs identified here also have previously defined role in pathogenesis of liver diseases. These data demonstrate the utility of cost-effective, animal-free, TF profiling assay for detecting DILI potential of drug candidates at early stages of drug development.

Helixconstraints and amino acid substitution in GLP-1 increase cAMP and insulin secretion but not beta-arrestin 2 signaling.

Glucagon-like peptide (GLP-1) is an endogenous hormone that induces insulin secretion from pancreatic islets and modified forms are used to treat diabetes mellitus type 2. Understanding how GLP-1 interacts with its receptor (GLP-1R) can potentially lead to more effective drugs. Modeling and NMR studies of the N-terminus of GLP-1 suggest a ß-turn between residues Glu9-Phe12 and a kinked alpha helix between Val16-Gly37. N-terminal turn constraints attenuated binding affinity and activity (compounds 1-8). Lys-Asp (i, i+4) crosslinks in the middle and at the C-terminus increased alpha helicity and cAMP stimulation without much effect on binding affinity or beta-arrestin 2 recruitment (compounds 9-18). Strategic positioning of helix-inducing constraints and amino acid substitutions (Tyr16, Ala22) increased peptide helicity and produced ten-fold higher cAMP potency (compounds 19-28) over GLP-1(7-37)-NH2. The most potent cAMP activator (compound 23) was also the most potent inducer of insulin secretion.

Short Hydrophobic Peptides with Cyclic Constraints Are Potent Glucagon-like Peptide-1 Receptor (GLP-1R) Agonists.

Cyclic constraints are incorporated into an 11-residue analogue of the N-terminus of glucagon-like peptide-1 (GLP-1) to investigate effects of structure on agonist activity. Cyclization through linking side chains of residues 2 and 5 or 5 and 9 produced agonists at nM concentrations in a cAMP assay. 2D NMR and CD spectra revealed an N-terminal ß-turn and a C-terminal helix that differentially influenced affinity and agonist potency. These structures can inform development of small molecule agonists of the GLP-1 receptor to treat type 2 diabetes.

Evaluation of a novel shape-based computational filter for lead evolution: application to thrombin inhibitors.

A novel shape-feature-based computational method is described and used to rapidly filter compound libraries. The computational model, built using three-dimensional conformations of active and inactive molecules, consists of a collection of whole molecule shapes and chemical feature positions that are ranked according to their correlation with activity. A small ensemble of these shapes and features is used to filter virtual compound libraries. The method is applied to two thrombin data sets and is shown to be efficient in identifying novel scaffolds with enhanced hit rates.

Comparing performance of computational tools for combinatorial library design.

In using computational tools for library design it is necessary to understand the performance and limitations of available methods. This letter reports systematic comparisons of applying ligand-based and structure-based tools across therapeutic project-derived data sets. Included are assessments of performance in real-world iterative design applications and the utility of target structural information. The results suggest that combining screening and target structure information is robust; further, a well-designed screening library can compensate for lacking structural information.

Application of Free-Wilson selectivity analysis for combinatorial library design.

In this chapter we present an application of in silico quantitative structure-activity relationship (QSAR) models to establish a new ligand-based computational approach for generating virtual libraries. The Free-Wilson methodology was applied to extract rules from two data sets containing compounds which were screened against either kinase or PDE gene family panels. The rules were used to make predictions for all compounds enumerated from their respective virtual libraries. We also demonstrate the construction of R-group selectivity profiles by deriving activity contributions against each protein target using the QSAR models. Such selectivity profiles were used together with protein structural information from X-ray data to provide a better understanding of the subtle selectivity relationships between kinase and PDE family members.

Transient inhibition of ATM kinase is sufficient to enhance cellular sensitivity to ionizing radiation.

In response to DNA damage, the ATM protein kinase activates signal transduction pathways essential for coordinating cell cycle progression with DNA repair. In the human disease ataxia-telangiectasia, mutation of the ATM gene results in multiple cellular defects, including enhanced sensitivity to ionizing radiation (IR). This phenotype highlights ATM as a potential target for novel inhibitors that could be used to enhance tumor cell sensitivity to radiotherapy. A targeted compound library was screened for potential inhibitors of the ATM kinase, and CP466722 was identified. The compound is nontoxic and does not inhibit phosphatidylinositol 3-kinase (PI3K) or PI3K-like protein kinase family members in cells. CP466722 inhibited cellular ATM-dependent phosphorylation events and disruption of ATM function resulted in characteristic cell cycle checkpoint defects. Inhibition of cellular ATM kinase activity was rapidly and completely reversed by removing CP466722. Interestingly, clonogenic survival assays showed that transient inhibition of ATM is sufficient to sensitize cells to IR and suggests that therapeutic radiosensitization may only require ATM inhibition for short periods of time. The ability of CP466722 to rapidly and reversibly regulate ATM activity provides a new tool to ask questions about ATM function that could not easily be addressed using genetic models or RNA interference technologies.

Predicting kinase selectivity profiles using Free-Wilson QSAR analysis.

Kinases are involved in a variety of diseases such as cancer, diabetes, and arthritis. In recent years, many kinase small molecule inhibitors have been developed as potential disease treatments. Despite the recent advances, selectivity remains one of the most challenging aspects in kinase inhibitor design. To interrogate kinase selectivity, a panel of 45 kinase assays has been developed in-house at Pfizer. Here we present an application of in silico quantitative structure activity relationship (QSAR) models to extract rules from this experimental screening data and make reliable selectivity profile predictions for all compounds enumerated from virtual libraries. We also propose the construction of R-group selectivity profiles by deriving their activity contribution against each kinase using QSAR models. Such selectivity profiles can be used to provide better understanding of subtle structure selectivity relationships during kinase inhibitor design.