Gotta love these straightforward studies that test a highly testable and simple hypothesis. This one reminds of me of the hypothesis that Kalash people in Northern Pakistan are a population partly derived of the ancestors of Alexander the Great's army. It turns out that these people do cluster out pretty strongly in at least one study. The evidence of a genetic contribution from Greeks however is mixed.
This study fails to support a similar hypothesis, that Romans contributed to the gene pool of a specific Chinese group. Here's a related news story. The Lemba, on the other hand are an example of this sort of story that has been supported by genetic evidence.
Testing the hypothesis of an ancient Roman soldier origin of the Liqian people in northwest China: a Y-chromosome perspective.
Zhou R, An L, Wang X, Shao W, Lin G, Yu W, Yi L, Xu S, Xu J, Xie X.
J Hum Genet. 2007 Jun 20; [Epub ahead of print]
Abstract: To test this hypothesis, 227 male individuals representing four Chinese populations were analyzed at 12 short tandem repeat (STR) loci and 12 single nucleotide polymorphisms (SNP). At the haplogroup levels, 77% Liqian Y chromosomes were restricted to East Asia. Principal component (PC) and multidimensional scaling (MDS) analysis suggests that the Liqians are closely related to Chinese populations, especially Han Chinese populations, whereas they greatly deviate from Central Asian and Western Eurasian populations. Further phylogenetic and admixture analysis confirmed that the Han Chinese contributed greatly to the Liqian gene pool. The Liqian and the Yugur people, regarded as kindred populations with common origins, present an underlying genetic difference in a median-joining network. Overall, a Roman mercenary origin could not be accepted as true according to paternal genetic variation, and the current Liqian population is more likely to be a subgroup of the Chinese majority Han.