Increasing the sample size is useful to get nice statistical properties out of noisy data. That doesn't mean you cannot learn anything from tiny samples, especially when the data isn't noisy (which is the case here, since they have the same set of genes, a lot of the noise is gone).