In the post-genomic era, there have been great advances in high-throughput experimental studies. These provide valuable observational data, however, integration of such data will provide a holistic understanding of cellular processes, which usually involve complex dynamic interaction of biological networks. The focus of our group is the development of computational approaches to integrate and analyze biological networks with the long-term applications in the discovery of drug targets and diagnostic markers.
We are interested in generation of comprehensive metabolic pathways by metabolic reconstruction approaches and including new reactions predicted for promiscuous enzymes. For reaction prediction, we will describe each reaction with a molecular descriptor, which can operate on substrate compounds to result in product compounds. A promiscuous reaction is predicted if the reaction operator results in a chemically valid product compounds. The resulting expanded metabolic reaction network will be investigated for the alternative metabolic pathways or for novel bio-transformation pathways. We are also interested in studying the evolutionary perspective of metabolic pathways by aligning these from various organisms and investigate emergence of their metabolic capabilities.
In approach to predict new reactions, a key assumption is that a promiscuous enzyme can accommodate compounds similar to its cognate substrate. To assess and further predict these, we will develop a combined sequence-structure and data mining approaches to predict ligand protein interactions. These prediction is based on reasoning that compounds sharing some similarity should also share targets, and targets sharing similar ligands should have similar binding sites. In machine learning approaches, we will derive conservation patterns between chemical substructures and binding site residues and implement them to predict ligand binding sites. For sequence/structure-based approach, we will develop computational strategies to predict meta-binding sites and consensus set of ligands combining structure-based ligand binding sites. In addition, these tools will also be used for identification of interaction modes of ligands at the families/sub-family level and will be used to suggest lead compound for drug discovery process. These predictions can be enhanced by incorporating knowledge of protein tertiary structures. Towards this, we will perform protein modeling using fragment assembly approach of Threading ASSembly and Refinement (TASSER) using constraints derived from templates with known structures.
- Pandit SB and Srinivasan N. 2003. Proteins: Structure function and genetics 52: 585-597. Survey for G-proteins in the prokaryotic genomes: prediction of functional roles based on classification. (Faculty of 1000: factor 6).
- Pandit SB, Zhang Y and Skolnick J. 2006. Biophysical Journal 91: 4180-4190. TASSER-Lite: An automated tool for protein comparative modeling.
- Pandit SB and Skolnick J. 2008. BMC Bioinformatics 9:531. Fr-TM-align : A new protein structural alignment method based on fragment alignments and the TM-score. (Highly accessed).
- Pandit SB and Skolnick J 2010. Proteins 78:2769-2780. TASSER_low-zsc: An approach to improve structure prediction using low z-score ranked templates.
- Carbonell P, Fichera D, Pandit SB, Faulon JL. 2012. BMC Syst Biol., 6:10 Enumerating metabolic pathways for the production of heterologous target chemicals in chassis organisms. (Highly accessed).