In experimental design, selecting the null hypothesis, H_0 is a critical first step. For that, a statistical test appropriate to the data is needed. Here, it is unclear what the nature of that data is. If, for example, each drug pair were tested once and the response variable Y is the percentage cell mortality, what is there to test? That 100% of one test showed 90% mortality? No statistics are involved in that case; it's just one observation.
On the other hand, if there were 10,000 observations and the mean cell death was 90%, the probability can be tested.