Roman slaves were primarily either fellow Romans or at least Italics, other Europeans, or natives from Roman SW Asia and coastal North Africa; Sub-Saharan Africans were a really small minority indeed.

I haven’t done that much research on chromosome deconvolution, but couldn’t the larger segments in North Africans be due to either continuous gene-flow, i.e. Egyptians, or multiple episodes of admixture, Maghrebians for example. Admixture in North Africa is much more consistent and dynamic than South Asia; North Africa’s geographic location between Tropical Africa, Europe, and SW Asia has facilitated a steady flow of gene-flow, even into the historic era.

The “North African” component is an admixed cluster, it’s typically moderately African admixed. For example, according to Treemix about 27% of the “NW African” cluster from Dodecad is derived from a population similar to modern West Africans. Interestingly, the “SW Asian” (18%) and some of the other southern Western Eurasian clusters (7%) also seem to contain traces of African-like gene-flow.

http://dienekes.blogspot.com/2012/03/using-treemix-with-admixture-components.html

could be roman. doubt it is as old as 4,000 years. recombination would have chopped stuff up a lot more. probably some really old stuff, “north african” component is african shifted last i checked. but not too much.

The question to be answered is: how recent is it? It exists in very similar levels in Berber and Arabic speaking North Africans of similar latitudes according to the ADMIXTURE results of Henn et al. 2012, despite the fact that extant Berber populations such as Tunisian Berbers were significantly isolated from other populations during the Islamic era (clear from the weird results of Tunisian Berbers at higher Ks). So the Negroid ancestry is probably mostly pre-Islamic in North Africa.

look the chromosome deconvolution. it looks pretty obviously mostly recent admixture to me (the ‘segments’ have not broken down yet, like they have in south asians).

Also, according to the ADMIXTURE results of Henn et al. 2012, Tunisian Berbers are very similar to the other tested apparently litoral North Africans (thus Saharawis, southern Moroccans and a portion of Egyptians, Libyans and Algerians are excluded) in their proportion of Negroid ancestry (Tunisian Berbers certainly have Negroid ancestry; it is clear from K=2):

http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1002397?imageURI=info:doi/10.1371/journal.pgen.1002397.g001

So, the Negroid ancestry in North Africans should be mostly pre-Islamic. It clearly follows a geographical pattern (increases from north to south) and there is no clear ethno-linguistic pattern according to the ADMIXTURE results of Henn et al. 2012.

There is no sufficient amount of evidence that would suggest that the bulk of African ancestry (which is in it self not a homogenous cluster) in North Africa dates back to the Islamic era. Are you implying slavery? North Africans are approximately 1/3 African, and there’s no reason not to assume that the majority of their African ancestry was autochthonous to the region prior to any Western Eurasian gene-flow into the continent. The fact that both Berbers and Egyptians speak African languages, carry African uniparental markers at moderate-predominant frequencies, and are consistently 1/3 autosomally African implies a much more ancient indigenous Tropical African presence in the Maghreb and most definitely the Nile Valley.

As you should already know, ROLLOFF functions most effectively with populations who’ve experienced single admixture events in the past (African-Americans for example), not populations who’ve either experienced admixture at various times in history, or exchange gene-flow at a consistent pace. The aforementioned therefore perfectly explains the results of Henn et al. 2012.

The 1,200 BP admixture date for the southern Moroccan population realistically corresponds to the most recent admixture event in which excess African admixture was assimilated into this particular Berber population. The southern Moroccan data set differs from other Maghrebi samples due to the presence of certain individuals, who are autosmally more African than the average North African. These outliers could be explained by one of two or both of the following scenerios, 1) they’re are the descendants of West African slaves or 2) they’re the descendants of indigenous Niger-Kordofanian speaking oasis dwellers who inhabited the region prior to the expansion of Berber speaking peoples in the past 2,000 years. Regardless, the initial admixture event between the African and Western Eurasian ancestors of modern Berbers likely predates the aforesaid admixture event by thousands of years.

The relatively recent date for “Nilotic admixture” in Egypt is easily explainable. Egypt, i.e. the Lower Nile Valley, is not separated from “Sub-Sahara Africa”, so therefore gene-flow between the Delta and Tropical Africa is consistent throughout history. The aforementioned reality skews the results of programs like ROLLOFF which only work effectively in particular simplified scenarios.

I would suggest you read the following study on mtDNA in North Africa, in particular Tunisia.

http://www.ncbi.nlm.nih.gov/pubmed/21082907

The inclusion of NE African samples will likely shed more light on the biological affinities of North Africans; all North Africans share a NE African signature of admixture, supporting the role of Afroasiatic speaking NE Africans in shaping the bioanthropological landscape of North Africa and SW Asia.