[Colloquium] TTIC Talks: Hetunandan Kamisetty, CMU

[Liv Leader]

When:     *Tuesday, March 15 @ 11*

Where:    *TTIC Conference Room #526*, 6045 S. Kenwood Ave, 5th Floor

Who:       *Hetunandan Kamisetty*, CMU

Title:       *Graphical Models and Games of Protein Sequence, Structure and
Evolution*

Proteins are the basic functional unit within the cell and
primarily display two kinds of variability: sequence and structure.
Understanding, and accurately accounting for this variability is crucial
to predict and engineer their behavior. I will describe how these sources
of variability can be modeled efficiently and effectively using undirected
Probabilistic Graphical Models (PGMs) and Graphical Games.

I will first describe GREMLIN, a method for learning generative
statistical models of evolutionarily related protein sequences. GREMLIN
uses convex optimization to learn undirected PGMs that capture
dependencies between both sequential and long-range pairs of
residues. The resulting generative models are well suited to designing new
proteins, and vastly out-perform existing methods based on Hidden
Markov Models and their variants. By effectively exploiting distributed
computing architecture, GREMLIN also presents a scalable alternative to
such approaches. I will then discuss GOBLIN, a method that uses PGMs
to model the structural variability of proteins. GOBLIN
uses variational inference to efficiently approximate the binding free
energy of bio-molecular interactions and estimate the effect of changes in a
protein
sequence on its interactions. GOBLIN is currently the most accurate
high-throughput method for this task.

Together, GREMLIN and GOBLIN provide a powerful framework for the modeling
of proteins and their interactions with other molecules. I will describe
an approach called GAMUT built on this framework. GAMUT models the
evolution of pathogenic proteins as they respond to a change in the
fitness landscape due to the introduction of a drug. GAMUT uses Graphical
Games to efficiently model the interaction between drugs and their protein
targets in a game-theoretic fashion. It accurately approximates the
Correlated Equilibria of such games using a novel method based on
relaxations to the marginal polytope and efficiently computes properties
of the equilibria of the game. GAMUT is a radically different approach to
drug design and promises to be an exciting way of designing and evaluating
future drug candidates.

Host: Jinbo Xu, j3xu at ttic.edu

-- 
Liv Leader
Faculty Services

Toyota Technological Institute
6045 S Kenwood Ave, #504
Chicago, IL 60637
Phone- (773) 834-2567
Fax-     (773) 834-9881
Email-  lleader at ttic.edu <jam at ttic.edu>
Web-   www.ttic.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20110310/b6bc9731/attachment.htm