On 2015 Nov 10, Peter Karp commented:

Although I like the notion that access statistics could be used to direct curation efforts for a given database, note there is a converse view of this problem: if I'm going to write programs that compute across an entire database, I want the entire database to be as accurate as possible, otherwise my large-scale compute will not return accurate results. For example, if I'm going to build a metabolic model from a database, I want the whole metabolic network within the database to be well curated, regardless of how often researchers happen to query a particular enzyme. If I'm going to search the whole human genome for a gene that has certain properties, I want the whole genome to be annotated/curated as accurately as possible, regardless of how often each gene is visited by the research community. Thus, taking shortcuts in curation could backfire when we want to query across an entire biological system.

On 2015 Nov 10, Peter Karp commented:

Although I like the notion that access statistics could be used to direct curation efforts for a given database, note there is a converse view of this problem: if I'm going to write programs that compute across an entire database, I want the entire database to be as accurate as possible, otherwise my large-scale compute will not return accurate results. For example, if I'm going to build a metabolic model from a database, I want the whole metabolic network within the database to be well curated, regardless of how often researchers happen to query a particular enzyme. If I'm going to search the whole human genome for a gene that has certain properties, I want the whole genome to be annotated/curated as accurately as possible, regardless of how often each gene is visited by the research community. Thus, taking shortcuts in curation could backfire when we want to query across an entire biological system.