Replication: a summary
To the extent possible, use biological replicates. Exactly what kind depends on the population towards which your inferences are aimed.
Technical replicates provide useful quality checks, but don’t reduce the variance as much as we would like.
Dye-swaps should be performed where appropriate.
Replicates make the use of formal statistical inference possible, though it is not necessarily easy to do so.