Development And Application of a Bayesian Framework For A Model of TRAIL-Mediated Apoptosis

Abstract:

Many biological networks are modeled using networks of ordinary differential equations (ODEs), but estimating the values of free parameters is a challenge particularly due to parameter nonidentifiability. We describe a guided Monte Carlo Markov Chain Bayesian search to explore a 78-dimensional parameter space for an ODE model of apoptosis, or programmed cell death. By calculating a Hessian matrix that captures local curvature in the landscape, we enable the search algorithm to sample flat regions and to move to lower positions in the landscape. Utilizing local curvature is more efficient as a method to search parameter space than a random walk alone.

This work is supervised by Prof. John Tsitsiklis and Prof. Peter Sorger.

Biography:

Hoda Eydgahi is Ph.D. candidate in the Department of Electrical Engineering and Computer Science at MIT. She received her M.S. from the same department in 2008. Prior to that, she obtained a B.S. in Biomedical Engineering from Virginia Commonwealth University in 2006. Her main research interests include applying probabilistic and machine learning techniques to biomedical fields.