Recent breakthroughs in high-throughput biology provide us with a wealth of
information derived
by novel experimental methods which include sequencing,
microarrays, protein-protein interaction
maps and other techniques to interrogate
biological cells. In this talk we outline a probabilistic framework
for representing
and fusing this information in the goal inferring the biological function of
genes from
genomic data supported by a set of partial models and databases.
The key computational ideas we use
are machine learning, graph representations
and probabilistic networks. The biological problems we will
address in the
talk include pathway inference, comparative genomics of closely related organisms,
and inference
of gene function using protein-protein interaction networks and
DNA chips.