Links for Additional Information

Srini Devadas, Ph.D.

Department Electrical Engineering & Computer Science
Assoc Dept Head, Professor of EECS

Room 38-427
(617) 253-0454 (Phone)


Srini Devadas has been with the faculty of Massachusetts Institute of Technology, Cambridge since 1988, and currently serves as Associate Head of the Department of Electrical Engineering and Computer Science. Devadas's research interests are in the areas of Computer-Aided Design (CAD) of VLSI computing systems, computer security, computer architecture, and computational biology. In the area of computational biology, he has done research with his students on protein and RNA structure prediction. He is also interested in hardware architectures that are tailored to solve computational problems arising in biology and biochemistry.

Research Summary

Proteins fill the majority of dry space in living cells and control most cellular functions through interactions with themselves and other biomolecules, as determined by their physical, chemical, and structural properties. While our understanding of genomes has grown dramatically over the years, interpreting the intertwined complexities found in proteomes remains a significant challenge.

Toward this goal, the Devadas group is interested in computational techniques for modeling protein interactions and structure based on statistical analysis and algorithmic optimization. Paramount to the approach is an emphasis on integrative techniques that can directly progress experimental research, as well as a dynamic view of protein structure populations.

Ensemble-based structure modeling

Determining the physical structure of a protein or protein complex is often one the best ways to understand the precise mechanism for protein interaction. Experimental techniques for doing so often incorporate software structure modeling tools that use available information, such as genetic sequence data, to predict conformational motifs and guide experimentation.

Until now, most computational tools have centered on predicting a single protein structure from a single amino acid sequence via numerical optimization. The Devadas group has instead developed techniques for the prediction of a Boltzmann ensemble of protein structures, that is, the statistical mechanical set of all likely protein conformations that may exist at once within a cell. At the same time, this approach uses a hypothesis-driven philosophy, which allows experimental knowledge (not simply sequence information) to be incorporated into the modeling algorithms, allowing for iterative experimental study.

For transmembrane β-barrel proteins, and important class of proteins that can act as a gatekeeper to the cell, this tool has been shown to better predict conformational motifs, while also enabling bulk population property prediction.

Ensemble-based sequence/structure modeling

Building upon the hypothesis-driven approach for Boltzmann ensemble structure prediction, the group has also developed algorithms for investigating the relationship between sequence and structure in both proteins and non-coding RNA (which can often interact with proteins). First, an algorithm has been devised which simultaneously folds and aligns two protein sequences, enabling evolutionary information to improve the prediction of structural ensembles, and conversely enabling structural ensemble prediction information to improve the alignment of homologous proteins.

Second, the structure of non-coding RNA molecules, and their dependence on specific nucleotide mutations were investigated. The group has presented an efficient algorithm for computing the grand canonical ensemble for RNAs — the Boltzmann set of all possible RNA structures for all possible sequences with k pointwise mutations. Such an analysis has been able to reveal pivotal nucleic acid positions controlling structural stability.

Selected Publications

  • Simultaneous Alignment and Folding of Protein Sequences. Jérôme Waldispühl, Charles W. O'Donnell, Sebastian Will, Srinivas Devadas, Rolf Backofen, Bonnie Berger. 13th Annual International Conference on Research in Computational Molecular Biology (RECOMB), Tucson, AZ, USA (May 2009)
  • Efficient Algorithms for Probing the RNA Mutational Landscape. Jérôme Waldispühl, Srinivas Devadas, Bonnie Berger, Peter Clote. PLoS Computational Biology 4(8):e1000124, (2008)
  • Modeling Ensembles of Transmembrane β-barrel Proteins. Jérôme Waldispühl*, Charles W. O'Donnell*, Srinivas Devadas, Peter Clote, Bonnie Berger (*authors equally contributed). Proteins: Structure, Function, and Bioinformatics Volume 71, Issue 3, pp. 1097-1112 (15 May 2008, digitally published 14 November 2007)
  • Learning Biophysically-Motivated Parameters for Alpha Helix Prediction. Blaise Gassend, Charles W. O'Donnell, William Thies, Andrew Lee, Marten van Dijk, Srinivas Devadas. BMC Bioinformatics Volume 8 (Supplement 5) p. S3 (24 May 2007)Last Updated: April 12, 2009