Y-chromosome 10 locus short tandem repeat haplotypes in a population sample from Sicily Italy
Ghiani et al. (2004)
Link to Abstract
Overall, results indicate Sicily is closest genetically to the mainland Italian population but also with evidence of a significant African component in the male gene pool.
That's from the Abstract and, as is often the case, greatly exaggerated and somewhat misleading. The body of the study reveals that the "African component" is specifically North African, and puts the use of the adjective "significant" into perspective:
Sicily and North West Africa share five of the seven-locus haplotypes.... Furthermore, these five haplotypes are not present in any other Italian population [20–23]. The shared five haplotypes represent 5% of the total Sicilian haplotypes [that's just 2.5% admixture]. These African haplotypes most probably were introduced into Sicily sometime between the 7th and 8th century, during the island’s domination by the Arab Empire.
The UPGMA tree (Fig. 2) visualizes the relationship of the populations of the Mediterranean Basin, Europe and North West Africa using pairwise distances. The North West African and Spanish (Spain and Basques) populations occupy an outgroup position within the tree, located some distance away with respect to the other European groups, which includes Sicily. Within the European cluster there is a tight grouping containing Sicily (South Italy), Italy (from all mainland), Germany, Holland, Hungary, Lombardy (North Italy) and Tuscany (Centre Italy). Sardinia lies in a separate branch at the edge of the cluster, and well way from Sicily.
An African contribution to the Sicilian gene pool gains support from several lines of evidence. ... Bernstein , Ragusa , Barrai ...agreed upon the fact that there has been a low but significant level of admixture with Africa. Using mtDNA haplotype frequency, Semino et al.  estimated that African gene flow into Sicily ranged between 10 and 34%.
The authors are a bit behind the times in terms of the research they've chosen to cite. The first three studies are based on the adaptive sickle cell gene, while the mtDNA study by Semino is from the 80s and problematic. Its results have not been duplicated by subsequent studies using more contemporary mtDNA sequencing methods. For example, Simoni et al. 2000 sampled a variety of Europeans, including Sicilians, and came to this contradictory conclusion:
Note that the analysis of molecular variance failed to identify any significant differences between northern and southern Europe; allele frequencies are roughly the same in the two regions.
Other studies like Vona et al. 2001 (TREE) and McEvoy et al. 2004 (PLOT) show a similar separation between Sicilian and North African/Middle Eastern mtDNA pools. And this, as well as the above Y-chromosome data, is reinforced by superior autosomal DNA testing. See, e.g., Kandil et al. 1999 (PLOT).
[NB: Excerpts from some of the cited studies and others can be viewed on this page.]