Yes, this is a good note for anyone interested in doing this. Other things to keep in mind, the paper source, while may be peer reviewed, most likely was not reproduced (a problem in the industry in general), so would be prudent to filter on # of publications supporting the SNP correlation. Also, the report will tell you relative risk, not absolute (e.g. 1.5x increased risk of getting breast cancer). In other words, your absolute risk may be 0.00001, so it can easily be 8x and it won't really matter. The most actionable items I've found for me personally were drug interactions due to genetics. Thanks for bringing up the caveats.