Hi,

I have a study design with 2 biological replicates where I want to know the differentially expressed genes in each of these across different time points. 

I want to contrast state DIAGNOSIS to the other states for each sample. What should be my design? I have tried a couple of ways to do this, going through forums but could not arrive at a contrast for all states e.g. A_stateDiagnosis_vs_RELAPSE1, B_stateDiagnosis_vs_RELAPSE3 etc.

Please suggest.

You can't do it per sample because you don't have replicates for this analysis: you just have a single observation per sample per state. You can do ~sample + state, where you would control for sample difference, and test for differences between the states.