To establish an extramurally-funded research program at the interface of chemical biology and informatics to investigate and solve problems in chemical biology.
Over the years doing research, I have realized that overarching goals of the research are to expand our knowledge of chemistry and biology through interdisciplinary collaboration and to encourage rapid medical improvements that can be rapidly translated to patient care. Working in close collaboration with other, multidisciplinary, research groups from different research areas has been an enriching experience. Most importantly, these projects have fostered a synergetic infosphere; my interactions with pharmacologists and synthetic chemists and clinicians involved in research have helped me see the field of research in a different light. The ultimate goal of my research would be to establish a symbiotic environment among the collaborating laboratories, investigating multidisciplinary approaches to harvest available chemical and biological data for drug discovery.
The research in my group is devoted to the development and application of computational methodologies to explore chemo-biological space and to use this knowledge for drug discovery and development. To achieve this goal we pursue research in three broad areas.
Cheminformatics and Information Harvesting
The prohibitive cost associated with high-throughput technology limited its application in the academic setting. Now that the technology is less expensive and more readily available, more and more academic labs have started producing large amounts of data. The recent explosion of publicly-available data depositories including biomedical texts documenting interactions between drugs, target proteins, biological events, and human pathologies provides novel opportunities to expand the repertoire of in silico drug discovery approaches. We propose well structured and comprehensive studies that initially make predictions using these separate lines of evidence coming from individual studies, and then combine those predictions and results following a rational and structured workflow, leading to better consensus predictions. In this project, we plan to explore concurrently and concordantly the diverse sources of data relevant to drug discovery. Using the available data, we will develop externally validated Quantitative Structure-Activity Relationship (QSAR) models to predict the biological profiles of diverse small molecules including marketed drugs. The overall goal of this project is to develop a drug discovery framework and to build and deliver efficient computational tools for rapid screening and profiling of small molecules. In addition, we will formulate a series of experimentally testable hypotheses for specific therapeutic areas such as cancer and other neglected diseases.
Ligand-based Drug Design
Ligand-based approach has typically relied on the description of compounds in the so called multidimensional chemistry space, i.e., compound representation by multiple chemical descriptors, with similarity searching and QSAR modeling being the most popular ligand based techniques. The main pursuit of this lab is to understand the relationship between chemical structure and its function and using that understanding to make testable predictions of which molecules are potential drug candidates. Our lab is focused, in part, on the development of general principles for modeling complex chemical-biological data where compounds are represented by conventional chemical descriptors.
We have a well established workflow and apply rigorous machine learning approaches to build robust and validated models. We employ multiple QSAR techniques based on combinatorial exploration of all possible pairs of descriptor sets coupled with various statistical data mining techniques to generate predictive models for different pharmacological and toxicological endpoints. Our current research interest is to develop mathematical/statistical models for predicting the human ADME/TOX for different class of drugs using methods such as quantitative structure pharmacokinetic relation (QSPKR), interspecies allometric scaling (AS) and in-vitro-in-vivo extrapolation (IVIVE). In order to do so we explore in-vitro, in-vivo pharmacokinetic data obtained from preclinical studies. We propose to build hybrid predictive models to predict in-vivo pharmacokinetic profile from in-vitro data. The models and methods developed would prove to be an important tool in drug development and will aid in understanding as well as predicting human pharmacokinetics.
Structure-based Drug Design
Understanding the three-dimensional aspects of drug-receptor interactions and their specificity at the molecular level has become a focal point in modern drug discovery. For many years computational medicinal chemists have attempted to develop computer tools that can evaluate protein structure and aid in the design of drugs with higher affinity and selectivity. The focus of our structure-based research is to develop a selectivity filter based on hybrid approach to screen ligands for structurally conserved protein targets as observed in many kinases and GPCRs. It is based on four-body statistical scoring function derived by combined application of the Delaunay tessellation of protein-ligand complexes and the definition of chemical atom types using the fundamental chemical concept of atomic electronegativity.
Physicochemical complementarity plays an important role in the process of molecular recognition and is widely adopted by medicinal chemists in structure-based drug design. Another focus of our study is to develop 3D ‘complementarity map’ based shape –property matching virtual screening platform. We propose complementary map-based virtual screening method could be both efficient and effective in virtual screening projects.
Our overall long-term goal is to establish and maintain a prototype of a Web-based tool that will support students and scientists by allowing them to access and share data for free for training and research purpose. We intend to share modeling software and specialized predictors with the research community through this Web-based portal.