So the best way to answer this question is to ask what is the variability in your system. This can be both technical and biological. Technical is a little more feasible to measure, but in practice it is really hard to measure either one of them . When it all comes down to it, we really have no idea how many replicates you need to do for your particular sample. 1,3 5, 7 ??? You need at least three to do a simple T-test, but beyond that it normally comes down to time and money. How much time and money do you want to invest and how much sample do you have. Based on past experience between 4-6 replicates is a good starting place. Usually we like to do a preliminary experiment to see how variability looks between conditions. After we do a preliminary experiment we can do a power analysis in MSStats to get an idea of how many replicates you need to see a certain fold change. Just a warning MSStats usually says you need far more replicates than you have time or money to acquire.
Here are some other resources on this topic
Chris Colangelo and I also hosted a variability workshop at ASMS in 2014. The slides can be found here