In Silico Prediction Method for Protein Asparagine Deamidation.

In silico prediction methods were developed to predict protein asparagine (Asn) deamidation. The method is based on understanding deamidation mechanism on structural level with machine learning. Our structure-based method is more accurate than the sequence-based method which is still widely used in protein engineering process. In addition, molecular dynamics simulation was applied to study the time occupancy of nucleophilic attack distance, which is hypothesized as the most important step toward the rate-limiting succinimide intermediate formation. A more accurate prediction method for distinguishing potentially liable amino acid residues would allow their elimination or reduction as early as possible in the drug discovery process. It is possible that such quantitative protein structure-property relationship tools can also be applied to other protein hotspot predictions.

Improving antibody thermostability based on statistical analysis of sequence and structural consensus data.

Background: The use of Monoclonal Antibodies (MAbs) as therapeutics has been increasing over the past 30 years due to their high specificity and strong affinity toward the target. One of the major challenges toward their use as drugs is their low thermostability, which impacts both efficacy as well as manufacturing and delivery. Methods: To aid the design of thermally more stable mutants, consensus sequence-based method has been widely used. These methods typically have a success rate of about 50% with maximum melting temperature increment ranging from 10 to 32°C. To improve the prediction performance, we have developed a new and fast MAbs specific method by adding a 3D structural layer to the consensus sequence method. This is done by analyzing the close-by residue pairs which are conserved in >800 MAbs' 3D structures. Results: Combining consensus sequence and structural residue pair covariance methods, we developed an in-house application for predicting human MAb thermostability to guide protein engineers to design stable molecules. Major advantage of this structural level assessment is in significantly reducing the false positives by almost half from the consensus sequence method alone. This application has shown success in designing MAb engineering panels in multiple biologics programs. Conclusions: Our data science-based method shows impacts in Mab engineering.

Protein asparagine deamidation prediction based on structures with machine learning methods.

Chemical stability is a major concern in the development of protein therapeutics due to its impact on both efficacy and safety. Protein "hotspots" are amino acid residues that are subject to various chemical modifications, including deamidation, isomerization, glycosylation, oxidation etc. A more accurate prediction method for potential hotspot residues would allow their elimination or reduction as early as possible in the drug discovery process. In this work, we focus on prediction models for asparagine (Asn) deamidation. Sequence-based prediction method simply identifies the NG motif (amino acid asparagine followed by a glycine) to be liable to deamidation. It still dominates deamidation evaluation process in most pharmaceutical setup due to its convenience. However, the simple sequence-based method is less accurate and often causes over-engineering a protein. We introduce structure-based prediction models by mining available experimental and structural data of deamidated proteins. Our training set contains 194 Asn residues from 25 proteins that all have available high-resolution crystal structures. Experimentally measured deamidation half-life of Asn in penta-peptides as well as 3D structure-based properties, such as solvent exposure, crystallographic B-factors, local secondary structure and dihedral angles etc., were used to train prediction models with several machine learning algorithms. The prediction tools were cross-validated as well as tested with an external test data set. The random forest model had high enrichment in ranking deamidated residues higher than non-deamidated residues while effectively eliminated false positive predictions. It is possible that such quantitative protein structure-function relationship tools can also be applied to other protein hotspot predictions. In addition, we extensively discussed metrics being used to evaluate the performance of predicting unbalanced data sets such as the deamidation case.

De novo prediction of p-glycoprotein-mediated efflux liability for druglike compounds.

P-glycoprotein (Pgp) is capable of recognizing and transporting a wide range of chemically diverse compounds in vivo. Overcoming Pgp-mediated efflux can represent a significant challenge when penetration into the central nervous system is required or within the context of developing anticancer therapies. While numerous in silico models have been developed to predict Pgp-mediated efflux, these models rely on training sets and are best suited to make interpolations. Therefore, it is desirable to develop ab initio models that can be used to predict efflux liabilities. Herein, we present a de novo method that can be used to predict Pgp-mediated efflux potential for druglike compounds. A model, which correlates the computed solvation free energy differences obtained in water and chloroform with Pgp-mediated efflux (in logarithmic scale), was successful in predicting Pgp efflux ratios for a wide range of chemically diverse compounds with a R(2) and root-mean-square error of 0.65 and 0.29, respectively.

Analysis of kinase inhibitor selectivity using a thermodynamics-based partition index.

In the quest for safe, efficacious kinase inhibitors as drugs, selectivity is often assessed early using kinase profiling panels. Here we present a selectivity index based on thermodynamics principles that can help in analysis of the resulting data. The "partition" selectivity index is easy to calculate and is applicable in certain situations where other widely used indices are not. It is uniquely useful in analysis of small, focused selectivity panel data frequently encountered in medicinal chemistry hit-to-lead and lead optimization. For larger "kinome" panels, the partition index allows assessment of selectivity relative to a kinase or multiple kinases of interest.

Design, synthesis, and biological evaluation of potent c-Met inhibitors.

c-Met is a receptor tyrosine kinase that plays a key role in several cellular processes but has also been found to be overexpressed and mutated in different human cancers. Consequently, targeting this enzyme has become an area of intense research in drug discovery. Our studies began with the design and synthesis of novel pyrimidone 7, which was found to be a potent c-Met inhibitor. Subsequent SAR studies identified 22 as a more potent analog, whereas an X-ray crystal structure of 7 bound to c-Met revealed an unexpected binding conformation. This latter finding led to the development of a new series that featured compounds that were more potent both in vitro and in vivo than 22 and also exhibited different binding conformations to c-Met. Novel c-Met inhibitors have been designed, developed, and found to be potent in vitro and in vivo.

Visual exploration of structure-activity relationship using maximum common framework.

To help tracking all molecules made in a typical medicinal chemistry project, we have developed an algorithm to generate a maximum common framework (MCF) hierarchy and an interactive tool for its visualization and analysis. By identifying all unique frameworks for a set of molecules and all molecules containing each framework, we were able to simplify the MCF hierarchy build up steps and, as a result, speed up the entire process significantly. By allowing compounds to be assigned to multiple MCFs, users can easily remove bad branching nodes and concentrate on interesting ones. MCF hierarchies provide an effective and intuitive visualization for tracking medicinal chemistry lead optimization projects. We will provide examples to illustrate its usefulness.

The discovery of 2-anilinothiazolones as 11beta-HSD1 inhibitors.

A series of 2-anilinothiazolones were prepared as inhibitors of 11beta-hydroxysteroid dehydrogenase type 1 (11beta-HSD1). The most potent compounds contained a 2-chloro or 2-fluoro group on the aniline ring with an isopropyl substituent on the 5-position of the thiazolone ring (compounds 2 and 3, respectively). The binding mode was determined through the X-ray co-crystal structure of the enzyme with compound 3. This compound was also approximately 70-fold selective over 11beta-HSD2 and was orally bioavailable in rat pharmacokinetic studies. However, compound 3 was >580-fold less active in the 11beta-HSD1 cell assay when tested in the presence of 3% human serum albumin.

Three-dimensional quantitative structure-activity relationships and activity predictions of human TRPV1 channel antagonists: comparative molecular field analysis and comparative molecular similarity index analysis of cinnamides.

3D-QSAR models for human TRPV1 channel antagonists were developed based on comparative molecular field analysis (CoMFA) and comparative molecular similarity analysis (CoMSIA), using a training set of 61 cinnamide TRPV1 antagonists and tested on an independent test set of 47 antagonists. Molecular alignment procedure included weights for both internal energy and atom-to-atom matching against a reference or probe. Sensitivity of results on partial charge assignments was explored using multiple charge sets. AM1-BCC charge assignments gave better results for both CoMFA and CoMSIA models. For the best CoMFA model, the statistics are, r2 = 0.96, q2 = 0.58, n = 61 for the training set and r2 = 0.50, n = 47 for the test set. For the best CoMSIA model, the statistics are r2 = 0.95, q2 = 0.57, n = 61 for the training set and r2 = 0.48, n = 47 for the test set. These models are consistent with the proposed binding modes and interactions of known activators of the TRPV1 channel such as capsaicin, in a structural model of the TM3/4 helical region of TRPV1.

Improving Docking Accuracy through Molecular Mechanics Generalized Born Optimization and Scoring.

Docking methods are typically used within the biopharmaceutical industry for the challenging purposes of suggesting putative binding modes of new chemotypes and for virtual screening. When attempting to satisfy the far more simplistic yet fundamentally important goal of reproducing and identifying the correct binding mode from a cocrystal, all docking methods fail at a rather significant rate, demonstrating room for further improvement in docking methodology. We report a hierarchical method that yields results comparable to the industry-leading docking packages GOLD, Glide, and Surflex. By first using a fast, simple, well-established method, UCSF DOCK 4.0, to rigidly dock conformational ensembles, we successfully generate the correct binding mode in all but 4 of a standard, publicly available set of 79 cocrystals from the PDB. Among these 4 failures (1glq, 1tmn, 1rds, and 8gch), all are highly flexible, highly charged, and not druglike. Subsequently, all resultant docking poses were optimized and scored in the protein with molecular mechanics, using a standard MMGB energy function. In total, this hierarchical method identified the correct binding in 71 of 79 cases (90%), an unprecedented level of accuracy on this highly benchmarked test set. Furthermore, the publicly available energy functions employ only physically based force fields without parameter fitting from this or any other docking test sets.

ADAAPT: Amgen's data access, analysis, and prediction tools.

The Amgen's Data Access Analysis Prediction Tools (ADAAPT) system is a desktop decision support tool developed to provide flexible access and analysis of chemical and biological data. The system is platform independent, adaptable, easily deployed, and scalable. It consists of four main modules: access, analysis, prediction, and tools. The access module contains numerous user interfaces designed to retrieve data easily. The analysis module provides standard computational tools to perform property calculation, QSAR/QSPR, and statistical analyses. The prediction module contains in-house models to calculate a drug-likeness score and absorption index. Finally, the tools module provides a wide array of features that are of general interest to our scientists.

FLAME: a program to flexibly align molecules.

Herein, we describe a method to flexibly align molecules (FLAME = FLexibly Align MolEcules). FLAME aligns two molecules by first finding maximum common pharmacophores between them using a genetic algorithm. The resulting alignments are then subjected to simultaneous optimizations of their internal energies and an alignment score. The utility of the method in pairwise alignment, multiple molecule flexible alignment, and database searching was examined. For pairwise alignment, two carboxypeptidase ligands (Protein Data Bank codes and ), two estrogen receptor ligands ( and ), and two thrombin ligands ( and ) were used as test sets. Alignments generated by FLAME starting from CONCORD structures compared very well to the X-ray structures (average root-mean-square deviation = 0.36 A) even without further minimization in the presence of the protein. For multiple flexible alignments, five structurally diverse D3 receptor ligands were used as a test set. The FLAME alignment automatically identified three common pharmacophores: a base, a hydrogen-bond acceptor, and a hydrophobe/aromatic ring. The best alignment was then used to search the MDDR database. The search results were compared to the results using atom pair and Daylight fingerprint similarity. A similar database search comparison was also performed using estrogen receptor modulators. In both cases, hits identified by FLAME were structurally more diverse compared to those from the atom pair and Daylight fingerprint methods.

Targeting a binding pocket within the trimer-of-hairpins: small-molecule inhibition of viral fusion.

Trimeric class I virus fusion proteins undergo a series of conformational rearrangements that leads to the association of C- and N-terminal heptad repeat domains in a "trimer-of-hairpins" structure, facilitating the apposition of viral and cellular membranes during fusion. This final fusion hairpin structure is sustained by protein-protein interactions, associations thought initially to be refractory to small-molecule inhibition because of the large surface area involved. By using a photoaffinity analog of a potent respiratory syncytial virus fusion inhibitor, we directly probed the interaction of the inhibitor with its fusion protein target. Studies have shown that these inhibitors bind within a hydrophobic cavity formed on the surface of the N-terminal heptad-repeat trimer. In the fusogenic state, this pocket is occupied by key amino acid residues from the C-terminal heptad repeat that stabilize the trimer-of-hairpins structure. The results indicate that a low-molecular-weight fusion inhibitor can interfere with the formation or consolidation of key structures within the hairpin moiety that are essential for membrane fusion. Because analogous cavities are present in many class I viruses, including HIV, these results demonstrate the feasibility of this approach as a strategy for drug discovery.

One-pot microwave assisted preparation of pyrazoloquinazolinone libraries.

The novel solution-phase synthesis of an array of biologically relevant pyrazoloquinazolinones in a simple microwave driven one pot procedure is revelaed. Transformations are carried out in good to excellent yield by condensation of alpha-cyano-ketones and 2-hydrazino-benzoic acids. Subsequent microwave irradiation affords pyrazoloquinazolinones with six points of potential diversification. The protocol described represents a very attractive solution phase procedure for the rapid generation of arrays of such functionalized cores, further demonstrating the growing importance of economic and enabling complexity generating chemistries in the lead discovery arena.

Knowledge-based approaches in the design and selection of compound libraries for drug discovery.

In the past decade, the pharmaceutical industry has realized the increasing significance of impacting the early phase hit-to-lead development in the drug discovery process. In particular, knowledge-based approaches emerged and evolved to address a multitude of issues such as absorption, distribution, metabolism and excretion (ADME), potency, toxicity and overall drugability. Each of these approaches seeks to bring together all relevant pieces of information and create a knowledge-oriented process to deploy such information in drug discovery. This review focuses on work relating to drugability, which aims at obtaining hits (or leads) that have enhanced likelihoods of leading to successful clinical candidates by medicinal chemistry efforts. The period covered in this review is from 1997 (since the publication of Lipinski's rule of 5) to March 2002.