Dear jxn,

HyPhy can do this fairly easily : you would need to define a non-reversible substitution model and then use Simulate to generate some data under that model. Were you looking to simulate nucleotide data?

Cheers,
Sergei

Dear jxn,

How do you wish to specify the input? A tree with branch lengths and a number of sites as well as substitution rates?

Cheers,
Sergei

Dear jxn,

Do you want the frequencies to be at equilibrium or not? If not, then branch lengths (in the traditional sense) are harder to compute...

Cheers,
Sergei

Dear jxn,

OK; stationarity makes things easier, but you have 3 constraints on 12 rate parameters to make sure that f_A, f_C, f_G and f_T you stipulate are the stationary distribution.

One needs to convert branch lengths (expected subs/site) into model parameters to simulate the data. I'll put something together tonight.

Cheers,
Sergei