# HMM depmixS4 using a vector of known states to fit model

I am using the depmixS4 package to fit HMMs to RNAseq count data.
My workflow is as follows:
Stack reads into a 'stack' vector which looks like this:

stack
[1]  3  3  3  3  3  3  3  3  3  3  3  3  3  4  4  4  4  4  4  4  4  4  4  4  4
[26]  4  4  4  4  4  4  4  4  4  4  4  4  6  6  6  6  7  7  7  8  8  8 10 13 13
[51] 13 13 13 13 13 16 16 16 16 16 16 14 14 14 14 14 14 14 14 14 15 15 15 16 16
[76] 16 16 17 17 18 18 18 18 18 18 18 18 18 18 21 22 23 23 23 23 23 23 23 23 23


To accompany this, for the same length vector (i.e. same chromosomal region) I have the known (manually assigned) state of the signal for a 2 state model i.e.:

RealState
[1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[38] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[75] 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2


It was my understanding that you could use the depmix() function with the real state vector to fit the model you like e.g.

MODEL <- depmix(stack~1, data = data.frame(stack), nstates=2,family=poisson(), transition = ~ RealState)
fitted.mod.depmix <- fit(MODEL)
hmmtrack <- posterior(fitted.mod.depmix)\$state


However, this does not give me any diffence from running the arbitrarily defined function without providing the RealState despite the fact that this explicitly states there are regions where signal is predicted from the original model that is false.

What is the proper way to provide known states to make a model better suited to my data?

Thanks.

EDIT:

I want to use RealState to initialize the model, and then estimate the parameters of a hidden Markov model (with unknown states).