Optimization Strategies for Protein Structure Alignment
Advisor: Prof. Nick Sahinidis (Carnegie Mellon University, Pittsburgh, PA )
Proteins are found in abundance is nature and are important due to their large range of functionality and indispensibility to the living cell. The functionality of proteins is dependent on the various different shapes and 3-d conformation the proteins assume. Thus similarity within protein structures can be used to elucidate the functional relationships between them and to classify them into their fold families and domains. My research aims at developing fast and accurate algorithms for this structure comparison and alignment and develop a comprehensive tool to perform protein structure database search and classification. Our group has developed a state-of-the-art structure alignment tool, CMOS, which has been further improved by the incorporation of reduction schemes based on biological information. This has led to an order of magnitude improvement in the performance of CMOS. We have also developed a new alignment tool, SAS-Pro (submitted to Bioinformatics), based on derivative-free optimization techniques to solve the more general case of non-sequential protein structure alignment.
Modeling of Formation of Complex Nanostructures
Advisor: Prof. Anurag Mehra (Indian Institute of technology, Bombay)
Reverse micellar systems are a popular method for producing nanoparticles where the micelles act as size-controlling agents for the nanopartices. There are a variety of issues in the reverse micellar route of nanoparticle formation such as the instability of the micelles and the processes tending to size enlargement by coagulation and Ostwald ripening. In this project, we studied and modelled the different processes of size enhancment of nanoparticles, such as Ostwald ripening and coagulation, and established their role in the fornation of silver halide nanoparticles.
Application of control theory for high Quality of Service (QoS), low overheads, communication networks
Advisor: Prof. Kannan Moudgalya and Prof. Krithi Ramamritham (Indian Institute of Technology, Bombay)
Online data, marked by its dynamic and unpredictable nature, is a constant source of information required to take many important decisions. It thus becomes necessary to obtain the right data from the pool of online data, and at the right time, with lowest possible transmission cost and minimal loss of information. This project deals with certain control theoretic approaches where a proportional controller and LQ controller are used to design a mechanism to perform online data polling with improved quality of service. The proposed control framework demonstrated effective tracking of dynamic data, and opened a new paradigm for online data tracking. This work was presented at the 17th IFAC World Congress, July 2008. [paper]
Parameter Estimation of Kinetics of Crystallisation
Advisor: Prof. Joerg Raisch (Max Planck Institute, Magdeburg, Germany)
The project involved estimating the kinetic parameters of growth and birth rates in a preferential crystallizer for the separation of L and D enantiomers of Threonine. The estimation was performed using a sctochastic approach (Adaptive Simulated Annealing) as well as a deterministic approach (Levenberg - Marquardt Algorithm), on data available from the experimental setup at Max Planck Insitute, Magdeburg. The resulting extimates were employed in designing a control framework for the preferential crystallizer.