ABSTRACT: In this talk, I will first introduce the requisite background in biological terminology so that non-specialists can easily follow the problem being studied, which is to use statistical classification techniques to determine which parts of a genome (DNA sequence) correspond to genes, and which parts don't. The most popular methods, based on multi-step Markovian models, are reviewed, and then I introduce a new algorithm called the Mixed Memory Markov Model (4M) algorithm that seems to perform as well as, or better than, the existing algorithms, while at the same time being amenable to rigorous statistical analysis. Along the way I will introduce notions such as the Kullback-Leibler divergence rate between Markov chains, which extends the familiar Kullback-Leibler divergence to the case of dependent stochastic processes. The algorithm proposed here is tailored for prokaryotes (lower life forms), and the challenges involved in extending it to eukaryotes (higher life forms) are briefly discussed. BIO: Dr. M. Vidyasagar received all of his degrees in Electrical Engineering from the University of Wisconsin, Madison, including the Ph.D. in 1969. After teaching in Canada for twenty years, lastly at the University of Waterloo, he returned to Bangalore, India in 1989 to set up the Centre for Artificial Intelligence and Robotics under the Ministry of Defence, Government of India. In 2000 he moved to Tata Consultancy Services, India's first and largest software and services firm, to set up the Advanced Technology Center in Hyderabad. The ATC consists of around 100 persons working in various areas, such as computational biology, e-security, open source/Linux, and quantitative finance. Dr. Vidyasagar has received many honors and awards, including the "Distinguished Service Citation" from his alma mater (the University of Wisconsin), the 2000 IEEE Bode Lecture Prize, and the 2008 IEEE Control Systems Award. |