How large should I expect R squared to be?

R² quantifies goodness-of-fit, as a fraction between 0.0 and 1.0, so of course you want it to be large. But how large? There is no general answer for this. The only answer is: It depends. Getting a large value of R² depends on having chosen an appropriate model, and also having little biological variation or experimental scatter. But the amount of biological variation and experimental imprecision varies widely between systems. In some systems, you expect R² to exceed 0.98 and you’d be concerned about experimental problems if R² was as low as 0.95. In other systems, you may expect an R² of only 0.80 or so, and you’d be concerned about someone faking or clipping data if R² was as high as 0.95.