Recent Applications of Statistical Discrete Models and Inference in Biology

• Activity Number: 630
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Charles Lawrence*+ and Luan Lin and Bryce Richards
• Companies: Brown University and Mt. Sinai Med Center and Brown University
• Keywords: high-D inference ; RNA ; Centroid estimators ; information theory

Abstract

The nearly vacant posterior ensemble of RNA secondary structures: Using an algorithm that identifies the set of the most informative bases we show that nearly all of the Boltzmann weighted ensemble of RNA secondary structures, > 98%, is concentrated in tiny fraction, of the sample space of secondary structures, where RNA secondary structure is defined by the joint occurrence of all its base pairs. Because, an RNA secondary structure can be represented by a tree its graphical model permits exact computations. This space is covered by a small number of energy wells, each of which is defined by a small number of the base pairs. Consequently, for each energy-well there is a full probabilistic model: the RNA secondary structure model subject to a set of constraints on the presents or absences of base pairs. We employ centroid estimates from each of these mutually exclusive energy wells to more concisely represent this high-D posterior space. Because we return inferences on complete RNA structures even though the space occupied by these defining wells is sparse, the solutions are not and as a result this approach provides an alternative approach to high-D inference.