Computational Tool for the Analysis of Human DNA

Period of Performance: 05/22/1998 - 11/21/1998

Unknown

Phase 1 SBIR

Recipient Firm

Neomorphic Software, Inc.
Berkeley, CA 94710
Principal Investigator

Abstract

DESCRIPTION (Adapted from investigator's abstract): The primary goal of this application is to develop and market software, that will employ hidden Markov models to predict gene structure from human EST sequence, genomic sequence, or a combination of both. The PI intends to use an iterative refinement approach. This technique is loosely modeled after a maximization methodology called the EM algorithm. This application proposes two variations to that approach. First, the use of stochastic models of gene structure and Bayesain statistics to estimate meters. Second, the use of user involvement in the iteration process. In a strict EM approach the algorithm itself iterates until a convergent to a solution is reached. The PI proposes to have a user involved in each step of the iteration. The intent of this is to ensure that the algorithm can convergent on a more optimal solution. Whenever possible, this tool will also identify protein motifs and sequence similarity in other sequences. Finally, it proposes to provide a Java based graphical user interface to visualize and integrate the results of the gene structure analysis. PROPOSED COMMERCIAL APPLICATION: NOT AVAILABLE