Program in Mathematics and Molecular Biology

PMMB Members

Fellowships

ª Former PMMB Fellows in Member Labs

Links

Click here to return to PMMB home.

RAM SAMUDRALA

Large Scale Protein Structure Prediction Using Comparative Modeling and ab initio Techniques

The goal of this project is to provide as much structural information as possible for new sequences that are being determined, including modelling complete genomes of organisms, using ab initio and comparative modelling methods. These methodologies that were developed were tested in a bona-fide manner at the third meeting on the Critical Assessment of Protein Structure Prediction (CASP3) in December 1998.

The following tasks have been accomplished in this area:

  1. Enhancement of conditional probability discriminatory function to better discriminate native from non-native conformations. This includes using combination functions and devising a new non-local packing function. This was tested at CASP3 with positive results (see publication list).

  2. A database of incorrect protein conformations, called Decoys ìRî Us (http://dd.stanford.edu) was developed to aid in development of discriminatory functions.

  3. Enchancement of side chain generation algorithm to better position sidechains. This was tested at CASP3 with positive results (see publication list).

  4. Developed an exhaustive main chain sampling algorithm for loop prediction, in conjunction with the graph theoretic-approach. This was tested at CASP3 with positive results (see publication list). In addition, a semi-database approach based on fragments from a library of protein structures has been developed. This will be tested at CASP4.

  5. Developed a build-up procedure for fitting secondary structure onto lattice models. This was tested at CASP3 with positive results (see publication list).

  6. Various collaborations for structural and functional genomics studies, including predicting structures for ORFs in the yeast genome and guiding experimental studies using predicted structure.

Future plans include building large number of models for well-characterised families, improvement of ab initio technique by using a fragment based simulated annealing approach for predicting structure, and improvement of discriminatory functions.

 
Click here to return to PMMB home.