James Watson admixture claims are invalid: graphic evidence

Last month, newspapers gleefully circulated deCODE Genetics' claim "that 16% of [Watson's] genes are likely to have come from a black ancestor of African descent [and] a further 9% of Watson’s genes are likely to have come from an ancestor of Asian descent." This alleged result of deCODE's analysis of Watson's DNA is absurd on its face (especially considering what is known about Watson's ancestry).

So what's going on here (ignoring, for the moment, that estimates of this sort provided [as deCODE do] without information on, e.g., confidence intervals, are meaningless)?

The images above are comparisons between an individual of European ancestry (Brian Smolenyak) and, from left to right, James Watson, Craig Venter, and South African Bantus. These images were generated by the "compare me" feature at deCODE's deCODEme website (stills are from this video; the Watson and Venter comparisons start about 15 minutes in, with the Bantu comparison appearing around 14:00) . Teal indicates "low sharing"; kelly green indicates "medium sharing"; and red indicates "high sharing". Notice anything? The video's creator did:

I was rather startled when I compared my husband's genome to that of James Watson. My initial reaction was that they must be different species. Well, I went to one of my favorite gurus, Ann Turner, who explained:

" . . . Watson's genome is very incomplete, with "too many" homozygous sites. If they were actually reading both chromosomes, there should be a number of heterozygous sites (different alleles)."

So Watson's incomplete genome accounts for the almost total lack of sharing.

Since the widely reported admixture analysis was done using a tool on the same website, there can be little doubt the same incomplete Watson data was used in that analysis. Ignoring any potential methodological flaws, bad data clearly invalidates deCODE's analysis. Stefansson surely knows this (he hinted as much to the NYT), but evidently for business and/or ideological reasons he is only too happy to spread lies.

A few more deCODEme genome comparisons are available here (details). Notice that all the comparisons (between predominantly European individuals) look a lot like the Venter comparison above, and very unlike the Watson comparison, in the ratio of green to teal.