|
|
|
|
|
|
|
|
| Deepak Bandyopadhyay, Johnson & Johnson PR&D |
|
Deepak Bandyopadhyay is a postdoctoral researcher in Dimitris Agrafiotis’ group at Johnson & Johnson Pharmaceutical R&D in Exton, PA, where he develops algorithms for molecular modeling, conformational analysis and 3D database search and helps integrate them within the ABCD data warehousing software for drug discovery. His research expertise is in the application of computational geometry and data mining techniques to solve problems in computational biology and chemistry.
Deepak received his B.Tech. in Computer Science and Engineering from the Indian Institute of Technology, Delhi in 1999, and Ph.D. in Computer Science from the University of North Carolina at Chapel Hill in 2005. His initial research at UNC was in computer graphics, wherein he designed and built a system (Dynamic Shader Lamps) for projector-based virtual painting on a moving object, an important milestone in the field of augmented reality.
Deepak’s dissertation research was in the more theoretical discipline of computational geometry, under the guidance of Jack Snoeyink. He observed that conventional geometric algorithms that assume precise input data are not reliable for finding nearest neighbors in protein structure coordinates and other inherently imprecise data, and introduced a new technique, Almost-Delaunay Simplices, for improved analysis of such data. He demonstrated applications of this technique to design robust analyses of protein packing, secondary structure and conformational change, in collaboration with Alex Tropsha at UNC’s School of Pharmacy. He also released an open-source software package (AlmDel), allowing other researchers to explore new applications in their own domains. In collaboration with data miners Luke Huan and Wei Wang he developed FFSM (Fast Frequent Subgraph Mining), the first practical program that can reliably find structural patterns occurring frequently within families of protein structures represented as graphs. Most recently, his work on structure-based function inference of proteins featured on the cover of Protein Science in June 2006.
|
|
A New Self-Organizing Algorithm for Molecular Alignment and Pharmacophore Development
Deepak Bandyopadhyay, Johnson & Johnson PR&D
Deepak Bandyopadhyay (speaker) and Dimitris K. Agrafiotis
We present a method for simultaneous 3D structure generation and pharmacophore-based alignment using a self-organizing algorithm called Stochastic Proximity Embedding (SPE). Current flexible molecular alignment methods either start from a single low-energy structure for each molecule and then tweak bonds or torsion angles, or choose from multiple conformations of each molecule. Methods that generate structures and align them iteratively (eg. Genetic Algorithms), are often slow.
In earlier work (2003), we used SPE to generate 3D structures by iteratively adjusting pairwise distances between atoms based on a set of rules, and showed that it samples conformational space better and runs faster than earlier programs. In this work, we run SPE on the entire ensemble of molecules to be aligned. Additional information on which atoms or groups of atoms in each molecule correspond to points of the pharmacophore can come from an automatically generated hypothesis or be specified manually. We add distance terms to SPE to bring pharmacophore points from different molecules closer, and also to line up normal/direction vectors associated with these points. We also permit individual atoms to be constrained to lie near external coordinates from a protein binding site. The 3D structures of each molecule in the resulting alignment are nearly correct if the pharmacophore hypothesis was chemically feasible; post-processing by BFGS minimization of the distance and energy functions further improves the structures and weeds out infeasible hypotheses.
The new tools can be used to develop and test 3D pharmacophores for a diverse set of known active compounds from a screening run, starting from only 1D correspondences between atoms derived from a pharmacophore hypothesis. The 3D pharmacophore extracted from a successful alignment can be used for 3D database searching.
|
|
|
|
|
|
|
|
|
|
|
