What you are describing is a "phenotype", namely a particular characteristic based on a gene or set of genes. In a word what the eye doctor told you is BS (and not boy scouts). Geneticists are reserved and will not ascribe a phenotype unless it can be reliably traced and reproduced either in a tractable model (we use mice) or demonstrated by genetic and mathematical models. In humans the phenotype most (greater than 90% consensus) will agree upon is blood type, as that has been experimentally proven and stands up to mathematical scrutiny. Assigning a phenotype to ancestors from west Africa (where the majority of African Americans originated) is a gross over reach of the available data, there are too many confounding factors to state a given population possesses a gene mutation / transcription / polymorphism that confers this quality. What happened was confirmation bias, the doctor saw something, and then latched on to it and just continued that observation until the finding was "confirmed."
All that said a colleague (and previous co-author) works at
Helix. He will state emphatically that the foundation is strong but concede the details are murky when it comes to this sort of thing. I knew guys that got 23andMe in the beta stages, and I was offered to participate in Helix's beta, but to me all this is just some nice hand waving, alchemy, and magic. There's science in it, and it will provide direction - vague incomplete direction - but any conclusions should be taken with a grain of salt. Certainly do not, under any circumstances, make any life changing decisions based on the results. rather submit to a proper genetic screening, with consultation by a medical geneticist if something looks odd.
BTW most of these tests look for collections of Single Nucleotide Polymorphisms (SNP's, SNiP's) that cluster in populations. Usually these gather together as they are passed on among generations and within geographic regions. The problem is these can statistically appear anywhere and in anyone. These companies select particular SNP's which appear to have originated in certain geographic regions and are present (or expressed) in contemporary local populations in order to assign heritage. However, as I stated these are not statistically bulletproof models and results can be widely askew due to very small variations. Helix claims to have a wider data set to which to compare. (This SNP appears often in a given place, you have it, so you must have ancestors from that place). They also have more Illumina machines in one building then some university systems have in the state.