My scientific research

The general area of my scientific research is computational biology, genomics, and proteomics. The goal is to elucidate processes responsible for DNA, RNA, protein and interactome structure, function, design, and evolution so as to understand (and reproduce by computing simulations) how an organism's genome specifies the behaviour and characteristics of the organism. The major things I've touched upon thusfar are listed below (for those interested in a more formal presentation, including an ordered list of publications, check out my CV).

I am a principal investigator (Associate Professor) at the University of Washington in Seattle interested in the following topics:

Specific areas of ongoing research are listed below. The work described in the publications is generally encapsulated into a variety of webservers/applications/services (links included) and downloadable software.

Application

Structural and functional studies of biologically important proteins, systems, and problems. Use the structure and function prediction tools developed by us to help guide experimentalists in manipulating proteins and extracting information about their function and structure in vivo, both at the single molecule as well as at the genomic/systems levels. Some key areas include work on therapeutic (inhibitor) discovery and nanobiotechnology. This work is usually done in collaboration with experimentalists. I list these papers first since they demonstrate a true application of the work we do. In many cases, these are prospective verification (i.e., a prediction is made before the answer is known and verified).

Systems

Application and integration of single molecule structure and function prediction techniques to whole genomes and proteomes in an integrated manner. Combine single molecule and genomic/proteomic data to to explore the relationships among the molecular and organismal (systems) worlds and create a comprehensive picture of the relationship between genotype and phenotype.

Interaction

Methods for predicting interactions between molecules.

Function

Generally applicable methods for predicting protein function from sequence and/or structure.

Structure

Protein structure from combining theory and experiment

Use the structure prediction methods described below with experimental data to produce better results.

De novo protein structure prediction

The basic paradigm is to sample the conformational space exhaustively (using lattice models) or semi-exhaustively (using discrete states for each amino acids with some stochastic search process) such that native-like conformations are observed. These conformations are selected using the all-atom based scoring functions described below. The methods have had good success in the CASP blind prediction experiments.

Comparative modelling of protein structure

This primarily uses a graph-theoretic clique-finding method to handle context-sensitivity issues in protein structures. The methods have had good success in the CASP blind prediction experiments.

Scoring/discriminatory functions for protein structure prediction

We primarily use an all-atom distance dependent conditional probability discriminatory function that is surprisingly accurate at selecting correct from incorrect protein conformations. It is used both for ab initio prediction and comparative modelling. We also use a number of other scoring functions as filters, and also develop databases of incorrect conformations ("decoys") to help evaluate scoring functions.

Side chain prediction

There are two papers in this area. The first is a work on exactly what it is that primarily determines side chain conformational preferences in proteins. The main thrust here is the use of the discriminatory function to select the most probable side chain rotamers given a large number of possible conformations. The second paper compares different methods for side chain prediction.

Infrastructure

We prefer to make our clusters from cheap components that can be readily discarded, and prefer to completely decentralise our systems.

Ram Samudrala || me@ram.org