A good class separation metric was given by Slonim et al.:
P(g,c) = (u1 - u2)/(s1 + s2)
where
- c: a class vector
- g: an expression vector of a gene over n samples
- u1: mean expression level in class 1
- u2: mean expression level in class 2
- s1: standard deviation in class 1
- s2: standard deviation in class 2