Boston University has been awarded a grant to develop Rational Genomic Annotation Systems for integration, mining and modeling of biological data. The research, which includes scientists and students at Massachusetts Institute of Technology and Tufts University, is aimed at developing a systematic methodology for the integration of multiple sources of evidence involving both predictive inference and learning. The PIs propose to develop an information-integration framework based on probabilistic graphical models, with a modular interface that will support different data types. This research will facilitate the development of a new generation of extensible gene annotation systems that will provide a greater coverage and accuracy than existing systems. It will also produce new computational algorithms to support these novel forms of integration as well as a systematic new methodology for building novel probabilistic model integrators. This research builds on a collaborative effort of several investigators emphasizing interdisciplinary research and education, training of a new generation of leaders with expertise in both system biology and computing sciences. The infrastructure and techniques proposed are aimed to provide the basic tools for novel wide international community efforts involving both computational biologists and experimental scientists to produce and validate the functional role of thousands of newly sequenced genes.