David Hecht

QSAR using evolved neural networks for the inhibition of mutant PfDHFR by pyrimethamine derivatives

Quantitative structure-activity relationship (QSAR) models were developed for dihydrofolate reductase (DHFR) inhibition by pyrimethamine derivatives using small molecule descriptors derived from MOE and/or QikProp and linear or nonlinear modeling. During this analysis, the best QSAR models were identified when using MOE descriptors and nonlinear models (artificial neural networks) optimized by evolutionary computation. The resulting models can be used to identify key descriptors for DHFR inhibition and are useful for high-throughput screening of novel drug leads.

A Novel In Silico Approach to Drug Discovery via Computational Intelligence

A computational intelligence drug discovery platform is introduced as an innovative technology designed to accelerate high-throughput drug screening for generalized protein-targeted drug discovery. This technology results in collections of novel small molecule compounds that bind to protein targets as well as details on predicted binding modes and molecular interactions. The approach was tested on dihydrofolate reductase (DHFR) for novel antimalarial drug discovery; however, the methods developed can be applied broadly in early stage drug discovery and development. For this purpose, an initial fragment library was defined, and an automated fragment assembly algorithm was generated. These were combined with a computational intelligence screening tool for prescreening of compounds relative to DHFR inhibition. The entire method was assayed relative to spaces of known DHFR inhibitors and with chemical feasibility in mind, leading to experimental validation in future studies.

Exploiting Atropisomerism to Increase the Target Selectivity of Kinase Inhibitors

Many biologically active molecules exist as rapidly interconverting atropisomeric mixtures. Whereas one atropisomer inhibits the desired target, the other can lead to off-target effects. Herein, we study atropisomerism as a possibility to improve the selectivities of kinase inhibitors through the synthesis of conformationally stable pyrrolopyrimidines. Each atropisomer was isolated by HPLC on a chiral stationary phase and subjected to inhibitor profiling across a panel of 18 tyrosine kinases. Notably different selectivity patterns between atropisomers were observed, as well as improved selectivity compared to a rapidly interconverting parent molecule. Computational docking studies then provided insights into the structure-based origins of these effects. This study is one of the first examples of the intentional preorganization of a promiscuous scaffold along an atropisomeric axis to increase target selectivity, and provides fundamental insights that may be applied to other atropisomeric target scaffolds.

Modeling the inhibition of quadruple mutant Plasmodium falciparum dihydrofolate reductase by pyrimethamine derivatives

Modeling studies were performed on known inhibitors of the quadruple mutant Plasmodium falciparum dihydrofolate reductase (DHFR). GOLD was used to dock 32 pyrimethamine derivatives into the active site of DHFR obtained from the x-ray crystal structure 1J3K.pdb. Several scoring functions were evaluated and the Molegro Protein-Ligand Interaction Score was determined to have one of the best correlation to experimental pKi. In conjunction with Protein-Ligand Interaction scores, predicted binding modes and key protein-ligand interactions were evaluated and analyzed in order to develop criteria for selecting compounds having a greater chance of activity versus resistant strains of Plasmodium falciparum. This methodology will be used in future studies for selection of compounds for focused screening libraries.

Computational Intelligence Methods for Docking Scores

Computer-aided drug design (CADD) methodologies have proven to be very effective, greatly enhancing the efficiency of small molecule drug discovery and development processes. These methods include quantitative structure- activity relationship and pharmacophore models, quantitative structure-property relationship models, as well as in silico docking studies. While docking studies very often correctly identify the binding mode of a ligand, they have reduced suc- cess in predicting binding affinities. Development of improved and more efficient strategies for scoring binding affinity is a very active area of research. Here we review the utility of computational intelligence approaches such as artificial neural networks, fuzzy logic, and evolutionary computation to the calculation of improved docking scores.

Applications of machine learning and computational intelligence to drug discovery and development

In silico modeling of ADMET property models with QSAR and QSPR models has proven to be an effective approach for increasing the efficiency of small molecule drug discovery and development processes. Development of new, improved models and techniques is currently an active area of research. In recent years, there has been growing interest in adapting tools and techniques from the fields of computational intelligence and machine learning for use in drug discovery and development. This report reviews some of the more popular applications. Drug Dev Res 72: 53–65, 2011.

CUREs in biochemistry—where we are and where we should go

Integration of research experience into classroom is an important and vital experience for all undergraduates. These course‐based undergraduate research experiences (CUREs) have grown from independent instructor lead projects to large consortium driven experiences. The impact and importance of CUREs on students at all levels in biochemistry was the focus of a National Science Foundation funded think tank. The state of biochemistry CUREs and suggestions for moving biochemistry forward as well as a practical guide (supplementary material) are reported here.

USING SCDL FOR INTEGRATING TOOLS AND DATA FOR COMPLEX BIOMEDICAL APPLICATIONS

Current bioinformatics tools or databases are very heterogeneous in terms of data formats, database schema, and terminologies. Additionally, most biomedical databases and analysis tools are scattered across different web sites making interoperability across such different services more difficult. It is desired that these diverse databases and analysis tools be normalized, integrated and encompassed with a semantic interface such that users of biological data and tools could communicate with the system in natural language and a workflow could be automatically generated and distributed into appropriate tools. In this paper, the BioSemantic System is presented to bridge complex biological/biomedical research problems and computational solutions via semantic computing. Due to the diversity of problems in various research fields, the semantic capability description language (SCDL) plays an important role as a common language and generic form for problem formalization. Several queries as well as their corresponding SCDL descriptions are provided as examples. For complex applications, multiple SCDL queries may be connected via control structures. For these cases, we present an algorithm to map a user request to one or more existing services if they exist.

Quantitative structure-property relationships for drug solubility prediction using evolved neural networks

Preclinical in vivo studies of small molecule compound libraries can be enhanced using a model of specific quantitative structure-property relationships. This may include toxicological or solubility measures such as prediction of drug solubility in mixtures of polyethylene glycol and/or water. Here we examine the utility of both multiple linear regressions and evolved neural networks for the prediction of drug solubility in aqueous solution. Initial results suggest that modeling requires compound libraries with high similarity. Clustering approaches can be used to group compounds by similarity with models built for each cluster. Linear and nonlinear models can be used for modeling, however evolved neural networks can be used to simultaneously reduce the feature space as well as optimize models for solubility prediction. With these approaches it is also possible to identify ldquohuman interpretablerdquo features from the best models that can be used by chemists during preclinical drug development.

Structural-based analysis of dihydrofolate reductase evolution

The evolution of dihydrofolate reductase (DHFR) was studied through a comprehensive structural-based analysis. An amino acid sequence alignment was generated from a superposition of experimentally determined X-ray crystal structures of wild-type (wt) DHFR from the Protein Data Bank (PDB). Using this structure-based alignment of DHFR, a metric was generated for the degree of conservation at each alignment site - not only in terms of amino acid residue, but also secondary structure, and residue class. A phylogenetic tree was generated using the alignment that compared favorably with the canonical phylogeny. This structure-based alignment was used to confirm that the degree of conservation of active-site residues in terms of both sequence as well as structure was significantly greater than non-active site residues. These results can be used in helping to understand the likely future evolution of DHFR in response to novel therapies.