The accelerated pace of biological research and the advent of genomics have resulted in the prodigious accumulation of data about genes and their relationship to disease. This information is scattered among millions of different records in sources such as Medline and OMIM. In order to rapidly and comprehensively summarize the biomedical literature, we used an automated approach to assemble disease gene co citation matrices from the titles, abstracts and MESH terms of over 11 million Medline records and normalized these gene disease relationships into rank order. We estimated the false negative rate for this tool to be 9.2%, whereas the false positive rate ranged from 22% to 30% depending on the strength of the association. The resulting database, termed MedGene, can generate lists of genes for use in high throughput screening experiments, can create disease specific micro arrays, and can sort the results of gene profiling data