First off, it's really important to know the difference between technical and biological replicates, since getting that right is the key to a good result. The best and most accurate way to handle the data is to fit a curve for each biological replicate on its own, and then average the IC₅₀ values you get.

Let's use your MTT test example. For each plate (which is one biological replicate), you'd first average the readings from your technical replicates—the multiple wells you have for each concentration. That gives you one full curve for that plate. You do this for all your separate experiments, so if you have three biological replicates, you'll end up with three separate curves.

Here’s the important part: you fit a curve to each of those three curves individually. This will give you three different IC₅₀ values. Then, you just average those three IC₅₀ values to get your final result.

If you average all your data into one "master" curve from the start, you're hiding the real-world messiness that happens between experiments. The error you'd calculate from that single, perfect-looking curve just shows how nicely the points fit that one line—not how much your results bounced around each time you actually ran the experiment. It can give you a false sense of confidence, making your result look super precise when it might not be.

But when you get three separate IC₅₀ values first, the variation between those numbers is what you really care about. That standard deviation shows you the true biological variation and answers the real question: "How much did my results naturally differ each time?" This gives you a much more honest and realistic measure of how precise your findings actually are.

Great answer by Jakub Jagielski

Hi Mingyen Sung To make research your IC50 calculation easier, you may use this online IC50 calculator. Here is the link https://mythreyaherbal.com/easy-ic50-calculator-for-herbal-drug-and-bioassay-studies/