Charles H. Reynolds

Lead Discovery Using Stochastic Cluster Analysis (SCA): A New Method for Clustering Structurally Similar Compounds

We have developed an algorithm that clusters structural databases using topological similarity. The first step in this procedure is to identify a set of probe structures that all fall outside a defined similarity score cutoff with respect to one another. This list of probes is then used to bin the remaining compounds in the database. In the last step, some housekeeping is performed to ensure that each compound in the dataset is either a probe or is contained in one and only one bin. We have applied this clustering method to a database of ∼27u2009000 compounds for which we have screening level biological data. Analysis of the resulting clusters shows that clusters defined by an active probe are much more likely to contain other active compounds than clusters defined by an inactive probe. Indeed, the incidence of active compounds in bins with active probes is anywhere from 6 to 10 times greater than the incidence of active compounds in the database as a whole. This results demonstrates the power of simple two-d...

Drug design : structure- and ligand-based approaches

Preface 1. Progress and issues for computationally guided lead discovery and optimization William L. Jorgensen Part I. Structural biology: 2. X-ray crystallography in the service of structure-based drug design Gregory A. Petsko and Dagmar Ringe 3. Fragment-based structure-guided drug discovery: strategy, process, and lessons from human protein kinases Stephen K. Burley, Gavin Hirst, Paul Sprengeler and Siegfried Reich 4. NMR in fragment-based drug discovery Christopher A. Lepre, Peter J. Connolly and Jonathan M. Moore Part II. Computational Chemistry Methodology: 5. Free-energy calculations in structure-based drug design Michael R. Shirts, David L. Mobley and Scott P. Brown 6. Studies of drug resistance and the dynamic behavior of HIV-1 protease through molecular dynamics simulations Fangyu Ding and Carlos Simmerling 7. Docking: a domesday report Martha S. Head 8. The role of quantum mechanics in structure-based drug design Kenneth M. Merz 9. Pharmacophore methods Steven L. Dixon 10. QSAR in drug discovery Alexander Tropsha 11. Predicting ADME properties in drug discovery William J. Egan Part III. Applications to drug discovery: 12. Computer-aided drug design: a practical guide to protein-structure-based modeling Charles H. Reynolds 13. Structure-based drug design case study: p38 Arthur M. Doweyko 14. Structure-based design of novel P2-P4 macrocyclic inhibitors of hepatitis C NS3/4A protease M. Katharine Holloway and Nigel J. Liverton 15. Purine nucleoside phosphorylases as targets for transition-state analog design Andrew S. Murkin and Vern L. Schramm 16. GPCR 3D modeling Frank U. Axe 17. Structure-based design of potent glycogen phosphorylase inhibitors Qiaolin Deng.

ESTIMATING LIPOPHILICITY USING THE GB/SA CONTINUUM SOLVATION MODEL: A DIRECT METHOD FOR COMPUTING PARTITION COEFFICIENTS

published in Advance ACS Absrmcrs, June 1. 1995. 0095-2338/95/1635-0738

Diversity and coverage of structural sublibraries selected Using the SAGE and SCA algorithms

09.00/0 the same molecule. The most significant problems with BLOGP are that it is built on an empirical expression which contains many highly correlated terms and that it includes some terms for which no clear mechanistic justification exists. These factors call the statistical significance of the BLOGP model into question. Recently CoMFA has also been used to derive lipophilicities directly from solute steric and electronic fields. For all their differences, these methods share the problems inherent in empirically derived models and may have limited validity outside their compound training sets. The most theoretically satisfying method for computing partition constants would be direct simulation of a solute in water and organic Once the free energy of solvation in aqueous and organic media have been determined, it is a simple matter to compute log P. This approach has been demonstrated with some success using molecular dynamics (MD)I4 and Monte Carlo (MC)I5-l7 methods. The advantages of all atom simulations are their generality and theoretical rigor. The disadvantage of using all atom MD or MC simulations to compute log P is that both are computationally very demanding and, therefore, impractical for computing the number of partition coefficients needed for a typical QSAR study. Another direct approach is to compute the free energy change for transfemng a solute from aqueous to organic solution using a continuum solvation model, such as the molecular mechanics based GB/SA model* or the quantum mechanical SM2 m0de1.I~ These methods are simpler and much more efficient than MD and MC simulations using explicit solvent and show promise for giving reasonable estimates of solution free The GB/SA method has also been parameterized to model one organic solvent, chloroform. This means that one can calculate the free energy associated with taking a compound from the gas phase to either water or chloroform. Using a simple thermochemical cycle this provides the free energy change from chloroform to water. The purpose of this paper is to demonstrate that chlorofondwater log P coefficients (log PCR) can be reliably estimated by computing this free energy change using the GB/SA solvation model.

Chemical information based scaling of molecular descriptors: A universal chemical scale for library design and analysis

It is often impractical to synthesize and test all compounds in a large exhaustive chemical library. Herein, we discuss rational approaches to selecting representative subsets of virtual libraries that help direct experimental synthetic efforts for diverse library design. We compare the performance of two stochastic sampling algorithms, Simulating Annealing Guided Evaluation (SAGE; Zheng, W.; Cho, S. J.; Waller, C. L.; Tropsha, A. J. Chem. Inf. Comput. Sci. 1999, 39, 738-746.) and Stochastic Cluster Analysis (SCA; Reynolds, C. H.; Druker, R.; Pfahler, L. B. Lead Discovery Using Stochastic Cluster Analysis (SCA): A New Method for Clustering Structurally Similar Compounds J. Chem. Inf. Comput. Sci. 1998, 38, 305-312.) for their ability to select both diverse and representative subsets of the entire chemical library space. The SAGE and SCA algorithms were compared using u- and s-optimal metrics as an independent assessment of diversity and coverage. This comparison showed that both algorithms were capable of generating sublibraries in descriptor space that are diverse and give reasonable coverage (i.e. are representative) of the original full library. Tests were carried out using simulated two-dimensional data sets and a 27 000 compound proprietary structural library as represented by computed Molconn-Z descriptors. One of the key observations from this work is that the algorithmically simple SCA method is capable of selecting subsets that are comparable to the more computationally intensive SAGE method.

Defining privileged reagents using subsimilarity comparison.

Scaling is a difficult issue for any analysis of chemical properties or molecular topology when disparate descriptors are involved. To compare properties across different data sets, a common scale must be defined. Using several publicly available databases (ACD, CMC, MDDR, and NCI) as a basis, we propose to define chemically meaningful scales for a number of molecular properties and topology descriptors. These chemically derived scaling functions have several advantages. First, it is possible to define chemically relevant scales, greatly simplifying similarity and diversity analyses across data sets. Second, this approach provides a convenient method for setting descriptor boundaries that define chemically reasonable topology spaces. For example, descriptors can be scaled so that compounds with little potential for biological activity, bioavailability, or other drug-like characteristics are easily identified as outliers. We have compiled scaling values for 314 molecular descriptors. In addition the 10th and 90th percentile values for each descriptor have been calculated for use in outlier filtering.

A Web-Based Chemoinformatics System for Drug Discovery

We have developed a new method for assigning a drug-like score to reagents. This algorithm uses topological torsion (TT) 2D descriptors to compute the subsimilarity of any given reagent to a substructural element of any compound in the CMC. The utility of this approach is demonstrated by scoring a test set of reagents derived from the Comprehensive Survey of Combinatorial Library Synthesis: 2000 (J. Comb. Chem.). R-groups were extracted from the most-active compounds found in each of the reviewed libraries, and the distribution of the subsimilarity scores for these monomers were compared to the ACD. This comparison showed a dramatic shift in the distribution of the JCC R-group subset toward higher subsimilarity scores in comparison to the entire ACD database. The ACD was also used to examine the relationship between molecular weight and various subsimilarity scoring algorithms. This analysis was used to derive a subsimilarity score that is less biased by molecular weight.

Combined Molecular Orbital and Group Additivity Approach for Modeling Thermochemical Properties: Application to Hydrazides

One of the key questions that must be addressed when implementing a chemoinformatics system is whether the tools will be designed for use by the expert user or by the bench scientist. This decision can impact not only the style of tools that are rolled out, but is also a factor in terms of how these tools are delivered to the end users. The system that we outline here was designed for use by the non-expert user. As such, the tools that we discuss are in many cases simplified versions of some common algorithms used in chemoinformatics. In addition, the focus is on how to distribute these tools using a web-services interface, which greatly simplifies delivering new protocols to the end user.

Chemistry, information and Frank: a tribute to Frank Brown

Heats of formation are often unavailable for compounds involved in synthesis, scale-up, or modeling of commercial products. This is true for agrochemicals developed at Rohm and Haas which are derived from hydrazides. I have computed heats of formation for five substituted hydrazides using AM1, HF/6-31G*, MP2/6-31 lG**, and MP4/6-31 lG**. The computed heats of formation were also used to derive a missing Benson group additivity equivalent for hydrazide. This equivalent makes it possible to compute heats of formation for hydrazides very easily using group additivity. Molecular orbital calculations are likely to be a valuable source for thermochemical information in the future, and this work shows their utility for extending group additivity into new classes of compounds previously inaccessible due to lack of experimental thermochemical data.

Drug Design: Contents

Sadly this is the 3rd tribute that we have published for founders of computational chemistry within the last year; the other two being Toshio Fujita and David Weininger. We lost Frank Brown when he died suddenly last year. He was a visionary. He was straightforward, open, often blunt, helpful, and carried a good sense of humor. He contributed substantially to our field. He made a difference. Frank Brown was an early founder of chemoinformatics, as he branded it, now known as cheminformatics. He headed the first dedicated cheminformatics group and published the first paper in the field. He made an impact on most likely everyone he met and then some. Also, he was my friend. We knew each other since we were both postdocs. We emailed some jokes just days before his demise and not long before that we discussed working together as consultants. He and I saw eye-to-eye on the integration of computational chemistry and cheminformatics into pharmaceutical research. However, I’ll keep my comments brief since the other six contributors to this piece knew him even better than I and all paint a vivid picture his scientific impact, his impact on the many people he knew, our friendships, and beyond all his loving family who placed a continual smile on his face. However, before we move on to the next author I’d like to note that these pieces don’t just bid farewell to our friends, colleagues, and comp chem founders, but are also meant to tell some history of comp chem. Our field is not so old, just a few generations. Although we’ve lost a some of the first generation, many of these founders are still with us. We still have the golden opportunity to learn from them. They started our field on its way and much of what we do as computational chemists is based on their insight. Also, some of what we routinely do is based on the shortcuts that were essential for their work 40 or 50 years ago. Had they the same information and computational resources that we have now it’s clear from their literature and from discussions with some of them that they would now do things differently. However, although a lot of substantial new advances have taken place, some aspects of our work have become rote and some of those shortcuts are still buried deep in our software and habits. Understanding history can give us a better idea of how to improve what we do. As a timely note, there will be a symposium in Franks honor Monday August 20, 2018 at the American Chemical Society’s 256th National Meeting in Boston, MA.