Searle Scholar Profile : Ram Samudrala (2002)

Modeling Genome Structure and Function

A fundamental biological challenge is to understand how the linear information in an organism’s genome is processed to produce the resulting behavior or phenotype. Genes are transcribed and translated into proteins that adopt three-dimensional conformations. Evolutionary processes ensure that a folded protein conformation interacts with its environment in a manner that is beneficial to the organism, using the protein to catalyse reactions, recognise cellular signals, build cellular structures, and to perform a host of other diverse biological functions.

Our research aims to understand these processes by constructing a structural and functional model for every tractable protein encoded by an organism’s genome, integrating theoretical predictions with experimental molecular, genomic and proteomic data. The goal of devising such a "genome prediction engine" is to develop a coherent picture of molecular and organismal structure, function, networks, and evolution within a fundamental scientific framework.

SPECIFIC AIMS

A. Construct models for all protein sequences with recognisable homology to known structures
in the protein databank to a high level of accuracy using comparative modelling techniques developed by us. The modelling is accomplished using a combination of graph-theoretic exhaustive enumeration methods, de novo simulation, and all-atom knowledge-based scoring functions, guided by information from publicly available sequence comparison programs.
B. Model sequences without recognisable homology to known structures by sampling protein conformations thoroughly and identifying conformations of biological interest. We use Monte Carlo and Genetic Algorithm searches with a combination of three move sets for sampling protein conformational space. Identification of native-like conformations is accomplished by combining a variety of all-atom based hierarchical filters that screen candidate structures based on pairwise atomic preferences, secondary structure, contact order, compactness and hydrophobicity.
C. Predict function using the resulting models with the aid of available experimental information. Large scale structure comparisons and microenvironment analyses, in conjunction with sequence-based methods and available experimental data, will be made to annotate functions for all the tractable sequences.
D. Develop confidence assessments to evaluate model quality for a given prediction. We will accomplish this through statistical analyses of the performance of our methods on proteins with known structures and functions, by comparing the predictions made for homologues and examining the consensus among them, and by comparing our predictions to those obtained from methods developed by other researchers, including publicly available webservers.
E. Publish the resulting information in an integrated manner so that it is useful for biologists to pose and answer precise scientific questions about organismal and systems biology. We will combine our predictions with experimental structural and functional Samudrala Research Plan 2 genomic/proteomic data via the use of database driven interfaces on the world wide web to accomplish this aim. A preliminary version of this effort can be found at the Bioverse data and webserver: .

RESEARCH BACKGROUND

Our research has focused on the prediction of protein structure and function, both at the molecular and genomic levels. We have devised graph theory-based approaches to predict short segments (loops) and side chain conformations of proteins, and exhaustive and semiexhaustive search methods to explore protein conformational space such that native-like conformations are sampled. We also use the graph theory and de novo simulation approaches to handle the alignment and refinement problems in comparative modelling of protein structure. We have successfully used all-atom conditional probability based scoring functions to select biologically relevant conformations from such samples. These methods were used successfully in blind prediction experiments at the first four Critical Assessment of protein Structure Prediction (CASP) meetings, and demonstrated significant progress in this field at the latest meeting (held in December 2000). Using the toolbox of methods developed (which are publicly available on our webserver and used by other biologists worldwide), predicted structure has been used to predict function, and to guide experimental work. The methods are currently being made more robust and automatic and being applied to entire genomes.

SIGNIFICANCE

We expect that the biological role of every protein sequence can eventually be deduced from its three-dimensional structure in the context of its environment in the cell. This information will enable us to probe that organism’s cellular pathways with an exquisite degree of sensitivity and also help us understand and treat infectious and inherited disease in an increasingly effcient and rational manner. The development of algorithms and tools to understand organismal genomes will have practical utility for pharmacogenomics and genetic engineering, and will be of use to the general research community to pose and answer ever more precise biological questions. Understanding organismal biology from a genomic perspective requires expertise in several scientific disciplines, including computing science, mathematics, physics, chemistry, and biology. The problems that need to be solved generally involve exploration of large search spaces and finding objects of interest within those spaces, as well as managing the large amount of data produced and making predictions from analysis of the data. Thus our research has significance in not only answering biological questions, but is also relevant for solving problems of a similar nature in other scientific disciplines.

LONG TERM GOALS

My research involves integrating knowledge from the fields of computing science, mathematics, biology, physics, and chemistry to:

Achieve better understanding of protein structure, protein function, and molecular evolution.
Analyse genomes and study interactions of individual genes and their corresponding proteins to understand and model their roles in infectious and inherited disease.
Use knowledge about the structure of proteins, protein-protein and protein-substrate interactions to model complete cellular pathways within an organism of interest.


	Ram Samudrala Assistant Professor Department of Microbiology Box 357242 University of Washington Seattle, WA 98195-7242 Voice: 206/732-6122 Fax: 206/732-6055 E-mail: ram@compbio.washington.edu Personal Homepage 2002 Searle Scholar