The number of people that must be sequenced to interpolate data on a disease depends on what the prevalence of that disease is in the population. A disease like xeroderma pigmentosum is quite rare and would require hundreds of thousands of genomes, while something more common like the BRCA1 gene mutation that leads to a susceptibility to some breast cancers would take 10 to 100 times fewer.
In reality however, each individual has unique mutations that may or may not effect their susceptibility to disease or ability to live a long life, and each new genome we sequence adds to our ability to correlate genes to disease. Indeed, individual actions of persons today could have far-reaching consequences for generations of people to come, particularly for their own progeny.
To have coverage so that 50% of criminals could be identified would require two components, a genome database and a relative database that distinguishes individuals related by blood rather than by marriage. The completeness of each database would determine the exact numbers needed.
That being said I don't think I would submit my personal genome to the database with the current uncertainty of personal protection and the state of the health care industry.