We need to perform miRNA expression profiling in our samples using affymetrix platform. Do the samples need to be done as technical replicates or biological replicates?
You want to use biological replicates to be sure to observe real effects. If you only do technical replicates, than you only test if your process (pipetting, sample preparation etc.) is consistant.
Both technical and biological replicates have their place. Technical replicates are especially useful for relatively new and untested platforms, to confirm reproducibility. Some would forgo them for established platforms like Affy, but it's never a bad idea to assess technical replicates when using a method new to the lab. For answering biological questions, though, technical replicates are uninformative, and biological replicates are needed, as Christian points out. How many you need depends on the effect you expect to see. Using pilot data from a small study, a biostatistician could perform a power calculation to determine how many samples you would need per group to achieve a given confidence level in a certain observed fold change.
I would do both, technical and biological replicates. Technical replicates are useful even when using a well established platform. You don’t need to replicate all your samples, 2 or 3 of them would be enough, just to address different days of hybridization, different technicians, etc. Biological replicates address the aim of your experiment and the number of them depends on the type of experiment. For example, if you are analyzing treated and untreated cell cultures, I would rather do at least 3 samples per group. For clinical samples you need to analyze a higher number of samples.