Detecting variants of unknown significance I have two groups of people, a healthy population which has carriers of disease and non-carriers, and I have a sick population.  I know the frequency of genetic variants in both groups.  I want to know if I can say anything about the variants with any statistical certainty and how I would do that.  
Variant      Frequency in Healthy Population    Frequency in Sick Population
X            30                                 25
Y            5                                  700
Z            600                                600

Each group (Healthy and Sick) has 1000 people in it.  Can I say anything about each variant?  
 A: I must admit that I have not a lot of experience with statistical genetics. And I dont fully understand what you mean by "carriers of disease and non-carriers". It is also not totally clear from your question what your goal is. The first thing you could do is a $\chi^{2}$-test to test whether the distributions of variants differ between sick and healthy people:
m <- matrix(c(30,5,600,25,700,600), ncol=3, nrow=2, byrow=T)
rownames(m) <- c("Healthy", "Sick")
colnames(m) <- c("X", "Y", "Z")

chisq.test(m)

    Pearson's Chi-squared test

data:  m
X-squared = 505.3131, df = 2, p-value < 2.2e-16

So we have strong evidence that the distribution of variants is different between sick and healthy people. If you can assume that the genetic variants are independent from each other, you could calculate the variant-specific odds ratios:
$$
OR_{X}=\frac{25\cdot 970}{975\cdot 30}=0.83\text{ (95%-CI: 0.46 to 1.47)}
$$
$$
OR_{Y}=\frac{700\cdot 995}{5\cdot 300}=464.33\text{ (95%-CI: 195.31 to 1457.65)}
$$
$$
OR_{Z}=\frac{600\cdot 400}{400\cdot 600}=1.00\text{ (95%-CI: 0.83 to 1.20)}
$$
So people carrying the variant $X$ are a bit less likely to be sick than people who don't carry the variant $X$. Further, carriers of variant $Y$ are 464 times more likely to be sick than non-carriers. Because there are exactly the same number of people carrying variant $Z$ among sick and healthy people, the odds ratio is 1.
You could also run a logistic regression with data on the indicidual level with disease status as outcome (1 = sick, 0 = healthy) and the variants as binary variables (1 = has variant xyz, 0 = doesn't have variant xyz). The exponentiated regression coefficients would then be the variant-specific odds ratios but adjusted for the other variants.
