There is a new paper onthe genetics of the Roma People (Gypsies), with emphasis in mtDNA:
Isabel Mendizabal et al., Reconstructing the Indian Origin and Dispersal of the European Roma: A Maternal Genetic Perspective. PLoS ONE, 2011. Open access.
The lineages of this European people of South Asian origins can be divided in a clearly South Asian component (M5a1 specially, also M18, M25 and M35b), a most likely West Asian component (X2 and J1 clades specially), a possibly Balcanic component (H7 and U3) and "others" (clearly European lineages). The exact apportions vary among populations as follows:
I think it is interesting demonstration of drift and founder effect how most pre-Europe "founder" lineages have vanished in most populations, specially in the rather well known "bottleneck" leading to non-Balcanic Roma: of all the Indian lineages almost only the largest one, M5a1, survives beyond that bottleneck. The same happens with West Asian X clades (barely surviving among Polish Roma). Instead other lineages have been amplified in destiny regions, no doubt by founder effect. This is particularly true for U3 but also H7 and some of the J1 clades and even an Indian lineage barely found in the Balcans (M18, which has thrived in Spain instead).
If you now compare Spanish or Lithuanian Roma with the Bulgaria 1 sample (probably the one best representing the ancestral Roma, at least in their "founder" lineages fraction), it is almost difficult to recognize much affinity. Only M5a1 remains as a clear link. The Bulgaria 3 sample (Vallachian Roma arrived to Bulgaria in the 19th century) is maybe a more clear ancestral link but still the differences are notorious.
The authors argue, on statistical methodology (table 3), on a Punjab origin for the Roma, which is consistent with their language. However, as Manju points out at his blog, this is not so consistent, it seems with their patrilineages, which are probably from SE India, lacking R1a1, the most common NW Indian lineage (and I would add R2, L, J2, etc., all of which were in India some 2000 years ago when the Roma exodus must have happened, at the earliest).
Actually the state of Orissa has also high statistical likelihood for the origin of the Roma people, being only second to Punjab by the authors' methodology. Also the main Roma founder mtDNA lineage, M5a1, is much more common in SE India than in the NW. This is also true for M18.
So it is surely worthy to consider, as Manju does, whether the proto-Roma are ultimately original from Eastern or SE India, although incorporating lineages maybe from the NW, such as M35b, the same that they later did in West Asia and Europe.