Why does differential expression testing need biological replicates?

Why

Imagine comparing a single treated patient specimen with a single control specimen. Tens of millions of reads measure those two specimens very precisely, but you still cannot tell whether a gene difference reflects treatment or normal donor variability. Independent donors or independently randomized culture units allow the statistical model to estimate biological dispersion, which drives interval width and uncertainty.

Technical replicates can be useful for instrument or pipeline repeatability and do not substitute for biological replication. Plan replicate counts using expected heterogeneity, effect sizes, power and study design, rather than treating any universal minimum as sufficient. In paired studies, collect matched observations from the same donors and encode donor effects appropriately.

Quality filtering and sequencing depth can improve count reliability, yet neither repairs a sample design without independent units. Keep inference aligned to the unit that received the intervention.