Human Genome Project databases present a confluence of interesting database challenges: rapid schema and data evolution, complex data entry and constraint management, and the need to integrate multiple data sources and software systems which range over a wide variety of models and formats. While these challenges are not necessarily unique to biological databases, their combination, intensity and complexity are unusual and make automated solutions imperative. We illustrate these problems in the context of the Philadelphia Genome Center for Human Chromosome 22, and describe a new approach to a solution for these problems, by means of a deductive language for expressing database transformations and constraints.