MendeleyPaper not really unique
A minor thing:
https://opensnp.org/search?utf8=%E2%9C%93&search=BRCA1&commit=Search
For some SNPs the Mendeley API returns a paper with a slightly different first author (in the above example: "X Huo" vs. "Xiang Huo", "X D Ma" vs. "X. D. Ma", "S J Ramus" vs "Susan J Ramus" etc.), the problem is that these papers have different reader counts, looks like the names etc. are manually changed over time.
The MendeleyPaper doesn't have a unique doi constraint which causes this (it only has a unique uuid constraint). However, just enforcing it would be boring, we wouldn't get the manual updates to the paper from the Mendeley database, we'd need a minor routine which checks for the paper using the doi and then updates all fields with the fields of the new copy from the Mendeley API. We'd also need a db migration which merges duplicate MendeleyPapers based on doi.