In the interests of giving credit to the resources geneticists find most useful, here are the the numbers of papers citing the most frequently cited links in Nature Genetics papers published in 2006 (volume 38). 108 of 184 papers cited links, totalling 531, with a mean of 2.9 per paper. The paper with the largest number of links was Park with 28.
9 NCBI Genbank which also has 261 linked nucleotide accessions (no repeat citations yet)
7 GEO with 25 microarray experiments and 5 microarray methods
4 ArrayExpress with 9 microarray experiments (one cited twice)
3 PDB with 3 protein accessions
5 Affymetrix (you have to double click to come back from this link)
In the examples above, the citing paper is linked by first listed corresponding author for convenience rather than in accordance with journal policy (which is to cite First Author et. al. or Corresponding Author and colleagues). By following the link, the full authorship can be seen. It is clear that accessions mandated by the journal predominate, but ideally, all publicly deposited resources should be fully cited by authors, and properly logged and counted by the journals.
Thanks to Irina Nudelman for help with analysis, any errors are my own fault – Myles Axton.