The short summary is -- apart from the genes related to the actual visible physical differences between "races", there are extremely few genes which reliably distinguish between "races". There are a few mutations (sickle-cell anemia, Tay-Sachs) which are much likelier to be found in one group than another, but which are still rare in all groups. But any gene which is widespread in one "race" and not related to visible racial characteristics is almost certain to be widespread in other races too.
Breeds of dogs differ by less, genetically, than the most pure-bred specimens of human races -- the human races are typically separated by a thousand or more generations, the dog breeds by a couple of hundred generations.
The proto-dog is the Dingo. If all dogs interbred randomly, it would only take only a few generations for them to all look Dingo-like. Dogs have been carefully bred by humans for thousands of years, more that just "a couple hundred generations." Humans have never been carefully bred to produce a human with certain characteristics (except for a few years in the 1940s). If you were to carefully breed humans to exaggerate certain characteristics, you could indeed breed humans with far different levels of distinction than exist in the present day as today's results are the product of fairly random breeding from a geographic area accessible to the couple. Indeed, as travel has become easier over the last 100 years and breeding is more common among people with different characteristcs, strict racial forms are less likely to be maintained.
Notice how you can't discuss this without sounding like a Nazi? Eugenics is the one scientific area of study still prohibited by the PC police.