Why are Dose response curves, as well as most biological graphs shown in semi log graphs and not normal graphs? What is the advantage of semi log graphs over regular graphs? I still haven't gotten an intuitive sense of the answer, can someone explain, maybe with a diagram?


The reason is to make it easier to determine a suitable dose from the graph. Normally, using a arithmetic scale, the response curve is a hyperbolic, with most of the information squished together in a small section of the graph. By using the log the function usually becomes somewhat sigmoid, with a relatively linear section in the transition from low-high response. This means that this dynamics in this section can be studied more carefully.

See e.g. this example (with data from https://www.idbs.com/media/396649/dose-response-comparison-example-hr.jpg), where I have reversed the x and y-axes, and used a secondary y-scale to show the log dose (so the graph is flipped compared to a normal dose-response curve). Here, the orange line shows a semi-log dose-response curve, and the blue line shows the same data using a normal arithmetic scale. Note that the orange line (log dose) should be read against the right scale and the "normal" dose against the left scale):

enter image description here

Here it is obvious that the normal scale doesn't give any possibility to interpret the region where the dose is effective, while this is obvious in the log graph.

The statistical modelling of these data is to some extent a separate issue though, where probit model or logit models along with data transformations are often used to analyse the data. Note also that the "usefulness" of the semilog-plot only lies in the visual presentation, and whether it will add anything depends on the dose series used in an experiment. If the doses used are not exponential (as in this example), the semilog-plot will not add much. In the statistical test of an effect the log dose is also usually not important, and I suspect (as someone not working in pharmacology) that the semilog plot is to some extent a historical artifact from the time before computer analysis of results, when the appropriate dose was determined in a more manual way (e.g. linear analysis of the middle portion of the semilog graph).

Example data:

dose (conc.)    response
0.00001         0.993
0.0001          1.02
0.001           3.12
0.01            6.36
0.1             31.97
1               51.2
10              75.96
100             89
1000            97.932
10000           98.8
  • 2
    $\begingroup$ Regarding your point about the historical background of the semi-log graphs, there's a few more factors. Namely, adding a competitive antagonist causes a parallel shift to the right in a semi-log graph - this is much simpler than the change in a graph that isn't semi-log. It also allows you to make a Schild plot comparing the amount of the parallel shift with the concentration of the antagonist, which can tell you the strength of the antagonist. $\endgroup$ – Jam Nov 14 '18 at 13:16
  • 2
    $\begingroup$ It also allows similar advantages for the analysis of noncompetitive antagonists and partial agonists. $\endgroup$ – Jam Nov 14 '18 at 13:16
  • $\begingroup$ @Jam Thanks for the additional background, and feel free to add to and edit my answer. Good to hear from somebody with more specific knowledge on dose-response curves than me. $\endgroup$ – fileunderwater Nov 14 '18 at 14:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.