The wikipedia article on Diabetes mellitus type 2 is a good place to start.
In the news (2019): 23andMe's Polygenic Risk Score for type-2 diabetes. A fine, short summary concluding that it is a good "engagement tool" but otherwise of no clinical use can be found here.
omim summarizes the latest research.
Numerous SNPs have each been associated with (slightly) increased risk for type-2 diabetes, but they only marginallly improve the odds of predicting whether an individual will get type-2 diabetes based on the traditional clinical characteristics combining age, sex and weight ([PMID 18694974]; see also ). Such SNPs include:
- 10 SNPs in all 10 regions found so far in multiple studies and meta-analyses, as known at the time of publication [PMID 17463248]:
- rs4402960, in the IGF2BP2 gene region, equivalent to rs1470579
- rs7754840, in CDKAL1 gene, equivalent to rs10946398 and rs4712523
- rs10811661, in an intergenic region on ch 9 near CDKN2A
- rs9300039, in an intergenic region on ch11, and 96% of time equivalent to rs1514823
- rs8050136, in the FTO gene region
- rs1801282 and rs17036314, in the PPARG gene
- rs13266634, in the SLC30A8 gene
- rs1111875 and rs7923837, near the HHEX gene, and equivalent to rs5015480
- rs7903146, in the TCF7L2 gene, and 85% of the time equivalent to rs7901695
- rs5219, in the KCNJ11 gene, and 99% of the time equivalent to rs5215
- Note that this paper reports that if you have the highest risk genotype for all 10 of these SNPs, you are estimated to be at double the risk for type-2 diabetes compared to the average person, whereas if you have the lowest risk genotype for every one of these 10 SNPs, you are estimated to be at half the risk.
- The 12 SNPs reported in a (2007) large, multi-lab Consortium study [PMID 17554300]:
- rs9465871, in the CDKAL1 gene fifth intron, like rs775840 and others
- rs4506565, in the TCF7L2 gene, and 92% of the time equivalent to rs7903146
- rs9939609, in the FTO gene, like rs9939609, rs7193144 and rs8050136
- Four SNPs discovered in a French population of diabetics [PMID 17293876], including:
- rs13266634, a SNP in the SLC30A8 gene
- SNPs in the SREBF1 gene (also known as RAI1) are associated with type-2 diabetes [PMID 17019602, PMID 18192539], including:
- A SNP in the SREBF2 gene is associated with risk of type-2 diabetes. [PMID 18192539]
- Two SNPs confirmed in a Japanese population:
- [PMID 17043802] Carrying two copies of a common variant of TCF7L2 doubles your chances of developing diabetes and puts you in a similar risk category to being clinically obese. A common variant of the gene increased the risk of developing diabetes by 50 per cent. Carrying two copies of the variant gene increased the risk another two fold. About 1 in 5 type-2 diabetics (18%) carry two copies of this variant compared to about 1 in 10 (8-11%) of the general population. In the population as a whole, the impact of this gene on the risk of developing diabetes is as big as the problem of being clinically obese (having a body mass index over 30). news summary. The SNP is either rs7903146 (IVS3C > T) or rs12255372
- DNA Direct's $500 Mail-In DNA test of TCF7L2 tests this SNP to derive one aspect of risk for type-2 diabetes 
But experts claim that promises of diagnostic tests for diabetes genes are misleading. And a reminder that a forecast of diabetes is not carved in stone.
[PMID 18162508] significantly associated with type-2 diabetes
- rs4402960, p = 0.00009;
- rs10811661, p = 0.0024;
- rs5219, p = 0.0034;
- rs1111875, p = 0.0064;
- rs13266634, p = 0.0073;
- rs7756992, p = 0.0363
in 1,630 Japanese subjects with type-2 diabetes and in 1,064 control subjects
[PMID 18984664] SNPs previously shown to be associated with higher type-2 diabetes risk were shown to also be associated with higher risk for gestational diabetes
[PMID 19002430] 6 SNPs previously shown to be associated with higher type-2 diabetes risk were shown to also be associated with higher risk for gestational diabetes specifically in Korean women
[PMID 18694974] 9 of 18 well-established genetic risk variants were associated with type-2 diabetes in a population-based study. Combining genetic variants has low predictive value for future type-2 diabetes at a population-based level. The genetic polymorphisms only marginally improved the prediction of type-2 diabetes beyond clinical characteristics.