The Zebrafish Database Project

Specific Aim 1. A Multi-media Database. We envision several kinds of data that will need to be included in these databases: Presently, we have examples of these three data types on-line. For example, photos illustrating the zebrafish developmental staging series are presently available as images on our existing WWW server [], as are various text files of methods, descriptions of mutants, addresses etc. Similarly, information about genetic linkage and the genetic map are available as graphical/spatial data on our ACeDB server (contact ernest@lenti.umn.edu).

Limitations of present systems. A major limitation of our current servers is that they can be accessed only by browsing, e.g. looking through a list of information like the names of cloned genes. Ultimately we will want to be able to make much more meaningful queries, like: What are all the genes known to be expressed by neurons in the nucleus of the posterior commissure at 24 hours of development, and what are the mutations known to perturb the expression of these genes? Such queries, that search across data in essentially unlimited combinations, require the full power of a relational database. Presently, such queries can be done only by consulting several separate data lists at different sites, and even if all the sites are known and consulted, it is unlikely that all the relevant data could be found because the mechanisms for searching these data lists are so primitive.

Fig 1. Relationship between neurons and gene expression patterns in 1-day zebrafish embryo. Left, image of neurons, including identified neurons of the nucleus of the posterior commissure (oval), stained with HNK-1 antibody [Wilson 90]. Right, expression pattern of the msxB homeobox gene in the same nucleus [Akimenko95].
Example of how the proposed system will work. We propose to develop a database that will facilitate complex interdependent searches. For example, to learn whether genes guide axons, we might want to know which developmental regulatory genes particular neurons express as they differentiate. Thus, descriptors of these neurons in one anatomical record (Fig 1, left) would need to be shared with the corresponding brain region in another record which shows patterns of gene expression, part of a completely different data set (Fig 1, right). We might then want to know more about the genes that these neurons express. From the expression pattern (Fig 1, right) we would want to find the gene sequence (Fig 2, upper right), information that could lead into public databases like GenBank. Knowing the location of the gene on the genetic map could lead to identification of mutations affecting the gene (Fig 2).
Fig 2. Relationships among msxB gene sequence (upper right, Akimenko95), location on the genetic map of chromosome I (left, Postlethwait94), and available mutant strains (lower right).
Descriptors of the mutants should identify the database containing images of the brain region of interest as it appears in the mutants (Fig 3).

Fig 3. Phenotype of an msxB deletion mutation. GAP-43 expression in the head of mutant (right) and wild-type sibling (left) embryos at 1 day of development.
In practice, all these databases should be linked in a manner allowing information to flow in any direction. For instance, the information flow outlined here could just as well proceed in the opposite direction, starting with a newly identified mutant phenotype, mapping it to a chromosomal location, identifying the corresponding gene, and retrieving the wild-type expression pattern of this gene and information about neurons in the affected region. Descriptors should also be shared with related data in human, mouse, and invertebrate databases.
The Zebrafish Database

Continue on to Specific Aim 2

Return to Table of Contents