As for your second question: biological replicates are fundamental to have good results (less false positives and better statistical power), but often times budget constraints and / or bad design decisions lead to studies without biological replication.

h.mon answered the second question well. On the first question: it depends? It's kinda hard to say based on what info you give. For instance, if there is a (reliable) genome available you can run alignment and perform DGE analysis, whereas if the reference genome is poor or nonexistent you would need to run a transcriptome assembly prior to alignment and analysis. Some folks do both (except maybe use a reference-guided assembly instead of running a de novo one.