Alex Tropsha, Ph.D., is an expert in the fields of computational chemistry, cheminformatics and structural bioinformatics who works to develop new methodologies and software tools for computer-assisted drug design. He is creating new approaches to protein 3D structure analysis and prediction based on the principles of statistical geometry. His particular expertise lies in the field of cheminformatics, a discipline where information and informatics methodologies are applied to storing, managing, exploring and exploiting chemical databases. In layman’s terms, cheminformatics combines chemistry and computer science to aid in the discovery of new drugs.
Tropsha has authored more than 190 peer-reviewed papers and 20 books and book chapters. He joined the School’s faculty in 1991 as an assistant professor and director of the Laboratory for Molecular Modeling. He was promoted to associate professor in 1997 and to full professor in 2004 and holds appointments as an adjunct professor in the UNC Department of Biomedical Engineering and in the Department of Computer Science and is a member of the UNC Lineberger Comprehensive Cancer Center. He was named as the K. H. Lee Distinguished Professor in 2008.
Early stages in modern drug discovery often involve screening small molecules for their effects on a selected protein target or a model of a biological pathway. In the past fifteen years, innovative technologies that enable rapid synthesis and high throughput screening of large libraries of compounds have been adopted in almost all major pharmaceutical and biotech companies and have also proliferated to academia leading to the unprecedented growth of available databases of biologically active compounds. Managing, understanding, analyzing and exploiting this data to enable rational design of new experiments requires skills and computational tools.
Tropsha’s Laboratory for Molecular Modeling conducts studies in the broad areas of computer-assisted drug design, cheminformatics and structural bioinformatics. The lab is generally interested in understanding the relationship between biomolecular structure and its function.
The lab is poised to develop
novel descriptors to characterize complex molecules,
novel tools for virtual screening of compound libraries and the design of novel compound and libraries with high expected hit rates, and
novel protocols for large scale cheminformatics computing and dissemination of the tools and target property predictors
The Laboratory for Molecular Modeling describes modern cheminformatics broadly as a chemocentric scientific disciplineencompassingthe creation, retrieval, management, visualization, modeling, discovery, and dissemination of chemical knowledge. QSAR modeling is the major cheminformatics approach to exploring and exploiting the dependency of chemical compound biological, toxicological, or other types of activities or properties on their molecular features. The lab has developed several major algorithms for QSAR modeling with the special emphasis on the general model validation protocols and published several seminal papers on general QSAR methodologies and on the matter of model validation, with one of them (J. Mol. Graphics Mod. 2002, 20, 269-276) recognized by Thomson ISI to be within 1% of most cited papers in the field. The best practices for developing statistically significant and externally predictive QSAR models are described in a recent review (Mol. Inf., 2010, 29, 476 – 488).
The rigor of the methodologies developed in the laboratory enable computational models that can be reliably used for virtual screening of compound libraries to identify putative bioactive compounds. The goal of such studies is to prioritize a small number of available chemicals for confirmative biological screening experiments with the expectation that a large fraction of computational hits is found active. The collaborative studies by the Laboratory for Molecular Modeling with several experimental groups afforded the discovery of novel anticonvulsants (J. Med. Chem. 2004, 47, 2356-64), anticancer compounds (J. Med. Chem., 2009, 52:4210-20; J Chem Inf Model. 2009, 49, 461-76), or GPCR binders (J Med Chem. 201011;53(21):7573-86).
Optimizing absorption, distribution, metabolism, excretion (ADME) and assuring lack of toxic side effects are critical for developing drug candidates as opposed to chemical tools to study biological phenomena. In silicomodeling of ADMET endpoints could provide a highly desirable alternative to animal testing via cost-efficient methodologies to eliminate compounds that are potentially unsafe or have undesired ADME properties in early phases of drug development. The Laboratory for Molecular Modeling has developed several models to predict protein binding (J Med Chem. 2006; 49,:7169-81), metabolic stability (J. Med. Chem. 2003, 46, 3013-3020), blood brain barrier permeability (Pharm.Res. 2008, 25(8):1902-14). In recent years, the lab has begun to develop new methodologies for chemical risk assessment that combine chemical descriptors and the results of in vitro assays treated as biological descriptors to arrive at hybrid chemical-biological models (Environ Health Perspect. 2009, 117(8):1257-64; Environ Health Perspect. 2011, 119(3):364-70).
Rapid growth of the protein structural databases such as Protein Data Bank (PDB) requires novel structural and functional classification schemes. There is also a related need for accurate annotation of orphan structures and sequences in the context of existing classifications. The chief hypothesis of the The Laboratory for Molecular Modeling is that structure and/or function specific amino acid residue motifs are encoded in residue packing patterns. These patterns can be discovered by the means of computational and statistical geometry analysis of protein structures, based on robust and objective definitions of nearest neighbor contacts between amino acid residues in folded protein structures. For the past 15 years, the lab has been involved in the application of computational geometry approaches to the analysis of protein structure-function relationships building models capable of predicting function of orphan proteins (J Comput Aided Mol Des. 2009, 23(11): 773-84; 785-97.
Cheminformatics plays a critical role in understanding the fundamental problem of structure-property relationships and therefore applies to almost any area of chemical and biological research. While it has been recognized as a distinct, impactful scientific discipline, there is a painful absence of cheminformatics tools in the public domain. Chembench is a publicly available cheminformatics portal (Bioinformatics, 2010. 26(23):3000-1). It provides cheminformaticians with the tools and data sets needed to create new models and provides chemical biologists and medicinal chemists with the tools to explore the predicted behavior of a broad range of compounds.
Biomedical Data Translator Technical Feasibility Assessment of Reasoning Tool (OT2)
Data Science: Infohub for Rare Diseases
PhD, biochemistry and pharmacology
Moscow State University
Advisor: Prof Lev S. Yaguzhinski
Thesis: Quantitative Structure-Activity Relationships for Muscarinic and Nicotinic Agonists and Antagonists
Moscow State University