Structure and Function of Proteins Using Computational Methods

Huzefa Rangwala

PhD Candidate
Department of Computer Science
University of Minnesota

Abstract

Proteins have a vast influence on the molecular machinery of life. Stunningly complex networks of proteins perform innumerable functions in every living cell. Knowing the three-dimensional structure of proteins is crucial to advances in biology, as this information provides insight into how proteins operate. For example, structural information enables function prediction, the identification of other interacting biomolecules (e.g., proteins, DNA and RNA), and the rational search for ligands that can be used to enhance or inhibit these interactions. As just a sampling of the implications of knowing protein structure, it can help cure the sick (develop better and more effective drugs for novel drug targets), feed the hungry (increase food production) and save the planet (serve in environmental remediation, and the development of biofuels).

In this talk I will highlight my work involving use of sequence information to characterize the structural and functional nature of proteins. In the first part of my talk, I will introduce the problem of remote homology detection and fold recognition, and discuss my contributions in the development of evolutionary-based string kernels. This work was extended to develop accurate multiclass classifiers that couple hierarchical information prevalent in these structural classification databases. In the second part of my talk I will discuss a novel local structure-based similarity measure (fRMSD) which is estimated from protein sequences. The predicted fRMSD scores have countless applications, and were shown to improve alignments between pairs of sequences having very low sequence identity. I will also demonstrate a generic protein sequence annotation toolkit (ProSAT), and biological web-services (MONSTER) that predict the local structural and functional properties of residues.

The above research has lead to contributions towards the overarching protein structure prediction problem i.e., determining the three-dimensional structure of a protein from a linear chain of amino acids. These methods have been successfully deployed in the biennial protein structure prediction competition called CASP. The talk will also provide future directions of research that I would like to pursue. One of the goals, include coupling structural bioinformatics with chemoinformatics in order to develop better virtual screening models, as well as provide an understanding of our life systems.

Speaker Bio

Huzefa Rangwala is a graduate student in the Computer Science department at the University of Minnesota, Twin Cities campus. He expects to graduate with a Ph.D. in Computer Science this year. His research interests lie in the areas of bioinformatics, machine learning and high-performance computing. His research work involves developing machine learning based methods for protein structure prediction. He has had internship experiences working at IBM developing life science applications for the Blue Gene/L, and has also been an instructor for a graduate level introductory bioinformatics class at the University of Minnesota.